1 ## Copyright (C) 1996-2025 The Squid Software Foundation and contributors
3 ## Squid software is distributed under GPLv2+ license and includes
4 ## contributions from numerous individuals and organizations.
5 ## Please see the COPYING and CONTRIBUTORS files for details.
10 ----------------------------
12 This is the documentation for the Squid configuration file.
13 This documentation can also be found online at:
14 http://www.squid-cache.org/Doc/config/
16 You may wish to look at the Squid home page and wiki for the
17 FAQ and other documentation:
18 http://www.squid-cache.org/
19 https://wiki.squid-cache.org/SquidFaq
20 https://wiki.squid-cache.org/ConfigExamples
22 This documentation shows what the defaults for various directives
23 happen to be. If you don't need to change the default, you should
24 leave the line out of your squid.conf in most cases.
26 In some cases "none" refers to no default setting at all,
27 while in other cases it refers to the value of the option
28 - the comments for that keyword indicate if this is the case.
33 Configuration options can be included using the "include" directive.
34 Include takes a list of files to include. Quoting and wildcards are
39 include /path/to/included/file/squid.acl.config
41 Includes can be nested up to a hard-coded depth of 16 levels.
42 This arbitrary restriction is to prevent recursive include references
43 from causing Squid entering an infinite loop whilst trying to load
46 Values with byte units
48 Squid accepts size units on some size related directives. All
49 such directives are documented with a default value displaying
52 Units accepted by Squid are:
54 KB - Kilobyte (2^10, 1'024 bytes)
55 MB - Megabyte (2^20, 1'048'576 bytes)
56 GB - Gigabyte (2^30, 1'073'741'824 bytes)
57 Squid does not yet support KiB, MiB, and GiB unit names.
59 Values with time units
61 Time-related directives marked with either "time-units" or
62 "time-units-small" accept a time unit. The supported time units are:
64 nanosecond (time-units-small only)
65 microsecond (time-units-small only)
74 year - 31557790080 milliseconds (just over 365 days)
77 Values with spaces, quotes, and other special characters
79 Squid supports directive parameters with spaces, quotes, and other
80 special characters. Surround such parameters with "double quotes". Use
81 the configuration_includes_quoted_values directive to enable or
84 Squid supports reading configuration option parameters from external
85 files using the syntax:
86 parameters("/path/filename")
88 acl allowlist dstdomain parameters("/etc/squid/allowlist.txt")
90 Conditional configuration
92 If-statements can be used to make configuration directives
96 ... regular configuration directives ...
98 ... regular configuration directives ...]
101 The else part is optional. The keywords "if", "else", and "endif"
102 must be typed on their own lines, as if they were regular
103 configuration directives.
105 NOTE: An else-if condition is not supported.
107 These individual conditions types are supported:
110 Always evaluates to true.
112 Always evaluates to false.
113 <integer> = <integer>
114 Equality comparison of two integer numbers.
119 The following SMP-related preprocessor macros can be used.
121 ${process_name} expands to the current Squid process "name"
122 (e.g., squid1, squid2, or cache1).
124 ${process_number} expands to the current Squid process
125 identifier, which is an integer number (e.g., 1, 2, 3) unique
126 across all Squid processes of the current service instance.
128 ${service_name} expands into the current Squid service instance
129 name identifier which is provided by -n on the command line.
133 Logformat macros can be used in many places outside of the logformat
134 directive. In theory, all of the logformat codes can be used as %macros,
135 where they are supported. In practice, a %macro expands as a dash (-) when
136 the transaction does not yet have enough information and a value is needed.
138 There is no definitive list of what tokens are available at the various
139 stages of the transaction.
141 And some information may already be available to Squid but not yet
142 committed where the macro expansion code can access it (report
143 such instances!). The macro will be expanded into a single dash
144 ('-') in such cases. Not all macros have been tested.
148 # options still not yet ported from 2.7 to 3.x
149 NAME: broken_vary_encoding
152 This option is not yet supported by Squid-3.
158 This option is not yet supported by Squid-3.
164 This option is not yet supported by Squid-3.
167 NAME: external_refresh_check
170 This option is not yet supported by Squid-3.
173 NAME: location_rewrite_program location_rewrite_access location_rewrite_children location_rewrite_concurrency
176 This option is not yet supported by Squid-3.
179 NAME: refresh_stale_hit
182 This option is not yet supported by Squid-3.
185 # Options removed in 7.x
189 Remove this line. Squid no longer supports this feature.
193 # Options removed in 6.x
197 Remove this line. Squid no longer supports this feature.
203 Remove this line. Squid no longer supports this feature.
206 NAME: announce_period
209 Remove this line. Squid no longer supports this feature.
215 Remove this line. Squid no longer supports this feature.
218 NAME: request_entities
221 Remove this line. Squid now accepts HTTP/1.1 requests with bodies.
222 To simplify UI and code, Squid rejects certain HTTP/1.0 requests with bodies.
225 # Options removed in 5.x
229 Remove this line. Squid no longer supports preferential treatment of DNS A records.
232 # Options removed in 4.x
233 NAME: cache_peer_domain cache_host_domain
236 Replace with dstdomain ACLs and cache_peer_access.
242 Remove this line. The behaviour enabled by this is no longer needed.
245 NAME: sslproxy_cafile
248 Remove this line. Use tls_outgoing_options cafile= instead.
251 NAME: sslproxy_capath
254 Remove this line. Use tls_outgoing_options capath= instead.
257 NAME: sslproxy_cipher
260 Remove this line. Use tls_outgoing_options cipher= instead.
263 NAME: sslproxy_client_certificate
266 Remove this line. Use tls_outgoing_options cert= instead.
269 NAME: sslproxy_client_key
272 Remove this line. Use tls_outgoing_options key= instead.
278 Remove this line. Use tls_outgoing_options flags= instead.
281 NAME: sslproxy_options
284 Remove this line. Use tls_outgoing_options options= instead.
287 NAME: sslproxy_version
290 Remove this line. Use tls_outgoing_options options= instead.
293 # Options removed in 3.5
294 NAME: hierarchy_stoplist
297 Remove this line. Use always_direct or cache_peer_access ACLs instead if you need to prevent cache_peer use.
300 # Options removed in 3.4
304 Remove this line. Use acls with access_log directives to control access logging
310 Remove this line. Use acls with icap_log directives to control icap logging
313 # Options Removed in 3.3
314 NAME: ignore_ims_on_miss
317 Remove this line. The HTTP/1.1 feature is now configured by 'cache_miss_revalidate'.
320 # Options Removed in 3.2
321 NAME: balance_on_multiple_ip
324 Remove this line. Squid performs a 'Happy Eyeballs' algorithm, this multiple-IP algorithm is not longer relevant.
327 NAME: chunked_request_body_max_size
330 Remove this line. Squid is now HTTP/1.1 compliant.
333 NAME: dns_v4_fallback
336 Remove this line. Squid performs a 'Happy Eyeballs' algorithm, the 'fallback' algorithm is no longer relevant.
339 NAME: emulate_httpd_log
342 Replace this with an access_log directive using the format 'common' or 'combined'.
348 Use a regular access.log with ACL limiting it to MISS events.
354 Remove this line. Configure FTP page display using the CSS controls in errorpages.css instead.
357 NAME: ignore_expect_100
360 Remove this line. The HTTP/1.1 feature is now fully supported by default.
366 Remove this option from your config. To log FQDN use %>A in the log format.
369 NAME: log_ip_on_direct
372 Remove this option from your config. To log server or peer names use %<A in the log format.
375 NAME: maximum_single_addr_tries
378 Replaced by connect_retries. The behaviour has changed, please read the documentation before altering.
381 NAME: referer_log referrer_log
384 Replace this with an access_log directive using the format 'referrer'.
390 Remove this line. The feature is supported by default in storage types where update is implemented.
393 NAME: url_rewrite_concurrency
396 Remove this line. Set the 'concurrency=' option of url_rewrite_children instead.
402 Replace this with an access_log directive using the format 'useragent'.
405 # Options Removed in 3.1
409 Remove this line. DNS is no longer tested on startup.
412 NAME: extension_methods
415 Remove this line. All valid methods for HTTP are accepted by default.
418 # 2.7 Options Removed/Replaced in 3.2
423 # 2.7 Options Removed/Replaced in 3.1
431 Remove this line. HTTP/1.1 is supported by default.
434 NAME: upgrade_http0.9
437 Remove this line. ICY/1.0 streaming protocol is supported by default.
440 NAME: zph_local zph_mode zph_option zph_parent zph_sibling
443 Alter these entries. Use the qos_flows directive instead.
446 # Options Removed in 3.0
450 Since squid-3.0 replace with request_header_access or reply_header_access
451 depending on whether you wish to match client requests or server replies.
454 NAME: httpd_accel_no_pmtu_disc
457 Since squid-3.0 use the 'disable-pmtu-discovery' flag on http_port instead.
460 NAME: wais_relay_host
463 Replace this line with 'cache_peer' configuration.
466 NAME: wais_relay_port
469 Replace this line with 'cache_peer' configuration.
474 -----------------------------------------------------------------------------
481 DEFAULT_DOC: SMP support disabled.
483 Number of main Squid processes or "workers" to fork and maintain.
484 0: "no daemon" mode, like running "squid -N ..."
485 1: "no SMP" mode, start one main Squid process daemon (default)
486 N: start N main Squid process daemons (i.e., SMP mode)
488 In SMP mode, each worker does nearly all what a single Squid daemon
489 does (e.g., listen on http_port and forward HTTP requests).
491 Changing the number of workers requires a restart: Squid warns about but
492 otherwise ignores attempts to change this setting via reconfiguration.
495 NAME: cpu_affinity_map
497 LOC: Config.cpuAffinityMap
499 DEFAULT_DOC: Let operating system decide.
501 Usage: cpu_affinity_map process_numbers=P1,P2,... cores=C1,C2,...
503 Sets 1:1 mapping between Squid processes and CPU cores. For example,
505 cpu_affinity_map process_numbers=1,2,3,4 cores=1,3,5,7
507 affects processes 1 through 4 only and places them on the first
508 four even cores, starting with core #1.
510 CPU cores are numbered starting from 1. Requires support for
511 sched_getaffinity(2) and sched_setaffinity(2) system calls.
513 Multiple cpu_affinity_map options are merged.
518 NAME: shared_memory_locking
521 LOC: Config.shmLocking
524 Whether to ensure that all required shared memory is available by
525 "locking" that shared memory into RAM when Squid starts. The
526 alternative is faster startup time followed by slightly slower
527 performance and, if not enough RAM is actually available during
528 runtime, mysterious crashes.
530 SMP Squid uses many shared memory segments. These segments are
531 brought into Squid memory space using an mmap(2) system call. During
532 Squid startup, the mmap() call often succeeds regardless of whether
533 the system has enough RAM. In general, Squid cannot tell whether the
534 kernel applies this "optimistic" memory allocation policy (but
535 popular modern kernels usually use it).
537 Later, if Squid attempts to actually access the mapped memory
538 regions beyond what the kernel is willing to allocate, the
539 "optimistic" kernel simply kills Squid kid with a SIGBUS signal.
540 Some of the memory limits enforced by the kernel are currently
541 poorly understood: We do not know how to detect and check them. This
542 option ensures that the mapped memory will be available.
544 This option may have a positive performance side-effect: Locking
545 memory at start avoids runtime paging I/O. Paging slows Squid down.
547 Locking memory may require a large enough RLIMIT_MEMLOCK OS limit,
548 CAP_IPC_LOCK capability, or equivalent.
551 NAME: hopeless_kid_revival_delay
554 LOC: Config.hopelessKidRevivalDelay
557 Normally, when a kid process dies, Squid immediately restarts the
558 kid. A kid experiencing frequent deaths is marked as "hopeless" for
559 the duration specified by this directive. Hopeless kids are not
560 automatically restarted.
562 Currently, zero values are not supported because they result in
563 misconfigured SMP Squid instances running forever, endlessly
564 restarting each dying kid. To effectively disable hopeless kids
565 revival, set the delay to a huge value (e.g., 1 year).
567 Reconfiguration also clears all hopeless kids designations, allowing
568 for manual revival of hopeless kids.
572 OPTIONS FOR AUTHENTICATION
573 -----------------------------------------------------------------------------
579 LOC: Auth::TheConfig.schemes
582 This is used to define parameters for the various authentication
583 schemes supported by Squid.
585 format: auth_param scheme parameter [setting]
587 The order in which authentication schemes are presented to the client is
588 dependent on the order the scheme first appears in config file. IE
589 has a bug (it's not RFC 2617 compliant) in that it will use the basic
590 scheme if basic is the first entry presented, even if more secure
591 schemes are presented. For now use the order in the recommended
592 settings section below. If other browsers have difficulties (don't
593 recognize the schemes offered even if you are using basic) either
594 put basic first, or disable the other schemes (by commenting out their
597 Once an authentication scheme is fully configured, it can only be
598 shutdown by shutting squid down and restarting. Changes can be made on
599 the fly and activated with a reconfigure. I.E. You can change to a
600 different helper, but not unconfigure the helper completely.
602 Please note that while this directive defines how Squid processes
603 authentication it does not automatically activate authentication. For a
604 given transaction, (re)authentication is requested in two primary cases
605 detailed below: initial authentication and re-authentication.
607 A client without credentials is requested to authenticate if one of the
608 following ACLs is evaluated by an http_access or adapted_http_access rule:
611 * proxy_auth_regex ACL
613 * external ACL with %ul logformat %code used in FORMAT parameters
614 * external ACL with %LOGIN macro used in FORMAT parameters;
615 this legacy macro currently behaves the same as %ul logformat %code
617 A client with credentials is requested to re-authenticate if http_access
618 or adapted_http_access denies its request _and_ the last evaluated ACL was
619 either proxy_auth, proxy_auth_regex, or an external
620 ACL with %ul or %LOGIN parameter (regardless of whether that last
621 evaluated ACL matched the denied request). Note that a max_user_ip ACL
622 does not have this effect: Requests denied after evaluating max_user_ip
623 trigger an HTTP 403 (Forbidden) response rather than re-authentication.
625 In both initial authentication and re-authentication cases, client access
626 is denied, typically with an HTTP 407 (Proxy Authentication Required) or
627 an HTTP 401 (Unauthorized) response.
629 WARNING: authentication can't be used in a transparently intercepting
630 proxy as the client then thinks it is talking to an origin server and
631 not the proxy. This is a limitation of bending the TCP/IP protocol to
632 transparently intercepting port 80, not a limitation in Squid.
633 Ports flagged 'transparent', 'intercept', or 'tproxy' have
634 authentication disabled.
636 === Parameters common to all schemes. ===
639 Specifies the command for the external authenticator.
641 By default, each authentication scheme is not used unless a
642 program is specified.
644 See https://wiki.squid-cache.org/Features/AddonHelpers for
645 more details on helper operations and creating your own.
648 Specifies a string to be append to request line format for
649 the authentication helper. "Quoted" format values may contain
650 spaces and logformat %macros. In theory, any logformat %macro
651 can be used. In practice, a %macro expands as a dash (-) if
652 the helper request is sent before the required macro
653 information is available to Squid.
655 By default, Squid uses request formats provided in
656 scheme-specific examples below (search for %credentials).
658 The expanded key_extras value is added to the Squid credentials
659 cache and, hence, will affect authentication. It can be used to
660 authenticate different users with identical user names (e.g.,
661 when user authentication depends on http_port).
663 Avoid adding frequently changing information to key_extras. For
664 example, if you add user source IP, and it changes frequently
665 in your environment, then max_user_ip ACL is going to treat
666 every user+IP combination as a unique "user", breaking the ACL
667 and wasting a lot of memory on those user records. It will also
668 force users to authenticate from scratch whenever their IP
672 Specifies the protection scope (aka realm name) which is to be
673 reported to the client for the authentication scheme. It is
674 commonly part of the text the user will see when prompted for
675 their username and password.
677 For Basic the default is "Squid proxy-caching web server".
678 For Digest there is no default, this parameter is mandatory.
679 For NTLM and Negotiate this parameter is ignored.
681 "children" numberofchildren [startup=N] [idle=N] [concurrency=N]
682 [queue-size=N] [on-persistent-overload=action]
683 [reservation-timeout=seconds]
685 The maximum number of authenticator processes to spawn. If
686 you start too few Squid will have to wait for them to process
687 a backlog of credential verifications, slowing it down. When
688 password verifications are done via a (slow) network you are
689 likely to need lots of authenticator processes.
691 The startup= and idle= options permit some skew in the exact
692 amount run. A minimum of startup=N will begin during startup
693 and reconfigure. Squid will start more in groups of up to
694 idle=N in an attempt to meet traffic needs and to keep idle=N
695 free above those traffic needs up to the maximum.
697 The concurrency= option sets the number of concurrent requests
698 the helper can process. The default of 0 is used for helpers
699 who only supports one request at a time. Setting this to a
700 number greater than 0 changes the protocol used to include a
701 channel ID field first on the request/response line, allowing
702 multiple requests to be sent to the same helper in parallel
703 without waiting for the response.
705 Concurrency must not be set unless it's known the helper
706 supports the input format with channel-ID fields.
708 The queue-size option sets the maximum number of queued
709 requests. A request is queued when no existing child can
710 accept it due to concurrency limit and no new child can be
711 started due to numberofchildren limit. The default maximum is
712 2*numberofchildren. Squid is allowed to temporarily exceed the
713 configured maximum, marking the affected helper as
714 "overloaded". If the helper overload lasts more than 3
715 minutes, the action prescribed by the on-persistent-overload
718 The on-persistent-overload=action option specifies Squid
719 reaction to a new helper request arriving when the helper
720 has been overloaded for more that 3 minutes already. The number
721 of queued requests determines whether the helper is overloaded
722 (see the queue-size option).
724 Two actions are supported:
726 die Squid worker quits. This is the default behavior.
728 ERR Squid treats the helper request as if it was
729 immediately submitted, and the helper immediately
730 replied with an ERR response. This action has no effect
731 on the already queued and in-progress helper requests.
733 NOTE: NTLM and Negotiate schemes do not support concurrency
734 in the Squid code module even though some helpers can.
736 The reservation-timeout=seconds option allows NTLM and Negotiate
737 helpers to forget about clients that abandon their in-progress
738 connection authentication without closing the connection. The
739 timeout is measured since the last helper response received by
740 Squid for the client. Fractional seconds are not supported.
742 After the timeout, the helper will be used for other clients if
743 there are no unreserved helpers available. In the latter case,
744 the old client attempt to resume authentication will not be
745 forwarded to the helper (and the client should open a new HTTP
746 connection and retry authentication from scratch).
748 By default, reservations do not expire and clients that keep
749 their connections open without completing authentication may
750 exhaust all NTLM and Negotiate helpers.
753 If you experience problems with PUT/POST requests when using
754 the NTLM or Negotiate schemes then you can try setting this
755 to off. This will cause Squid to forcibly close the connection
756 on the initial request where the browser asks which schemes
757 are supported by the proxy.
759 For Basic and Digest this parameter is ignored.
762 Useful for sending credentials to authentication backends that
763 expect UTF-8 encoding (e.g., LDAP).
765 When this option is enabled, Squid uses HTTP Accept-Language
766 request header to guess the received credentials encoding
767 (ISO-Latin-1, CP1251, or UTF-8) and then converts the first
768 two encodings into UTF-8.
770 When this option is disabled and by default, Squid sends
771 credentials in their original (i.e. received) encoding.
773 This parameter is only honored for Basic and Digest schemes.
774 For Basic, the entire username:password credentials are
775 checked and, if necessary, re-encoded. For Digest -- just the
776 username component. For NTLM and Negotiate schemes, this
777 parameter is ignored.
779 IF HAVE_AUTH_MODULE_BASIC
780 === Basic authentication parameters ===
782 "credentialsttl" timetolive
783 Specifies how long squid assumes an externally validated
784 username:password pair is valid for - in other words how
785 often the helper program is called for that user. Set this
786 low to force revalidation with short lived passwords.
788 NOTE: setting this high does not impact your susceptibility
789 to replay attacks unless you are using an one-time password
790 system (such as SecureID). If you are using such a system,
791 you will be vulnerable to replay attacks unless you also
792 use the max_user_ip ACL in an http_access rule.
794 "casesensitive" on|off
795 Specifies whether upper case letters in client-sent usernames are
796 preserved. By default and when explicitly set to "off", a username
797 extracted from Proxy-Authorization or Authorization request header is
798 forced to lower case before user credentials are checked or stored.
800 Most user databases are case insensitive, allowing the same username to be
801 spelled using both lower and upper case letters. For such databases, any
802 casesenstive setting should work, but forcing usernames to lower case may
803 still make a big difference for Squid internal caches like those used by
804 an external ACL with %un logformat code in FORMAT and a user_max_ip ACL.
806 When working with a case sensitive database, set casesensitive to "on".
808 Squid ACLs like proxy_auth are case-sensitive by default. An ACL using
809 upper case letters in user names (e.g., `acl badGuys proxy_auth Bob`)
810 will not match any user with Basic Authentication credentials unless
811 casesensitive is explicitly turned "on" (to preserve "Bob" username
812 instead of converting it to "bob" before the ACL is checked).
815 IF HAVE_AUTH_MODULE_DIGEST
816 === Digest authentication parameters ===
818 "nonce_garbage_interval" timeinterval
819 Specifies the interval that nonces that have been issued
820 to client_agent's are checked for validity.
822 "nonce_max_duration" timeinterval
823 Specifies the maximum length of time a given nonce will be
826 "nonce_max_count" number
827 Specifies the maximum number of times a given nonce can be
830 "nonce_strictness" on|off
831 Determines if squid requires strict increment-by-1 behavior
832 for nonce counts, or just incrementing (off - for use when
833 user agents generate nonce counts that occasionally miss 1
834 (ie, 1,2,4,6)). Default off.
836 "check_nonce_count" on|off
837 This directive if set to off can disable the nonce count check
838 completely to work around buggy digest qop implementations in
839 certain mainstream browser versions. Default on to check the
840 nonce count to protect from authentication replay attacks.
842 "post_workaround" on|off
843 This is a workaround to certain buggy browsers who send an
844 incorrect request digest in POST requests when reusing the
845 same nonce as acquired earlier on a GET request.
849 === Example Configuration ===
851 This configuration displays the recommended authentication scheme
852 order from most to least secure with recommended minimum configuration
853 settings for each scheme:
855 #auth_param negotiate program <uncomment and complete this line to activate>
856 #auth_param negotiate children 20 startup=0 idle=1
858 #auth_param digest program <uncomment and complete this line to activate>
859 #auth_param digest children 20 startup=0 idle=1
860 #auth_param digest realm Squid proxy-caching web server
861 #auth_param digest nonce_garbage_interval 5 minutes
862 #auth_param digest nonce_max_duration 30 minutes
863 #auth_param digest nonce_max_count 50
865 #auth_param ntlm program <uncomment and complete this line to activate>
866 #auth_param ntlm children 20 startup=0 idle=1
868 #auth_param basic program <uncomment and complete this line>
869 #auth_param basic children 5 startup=5 idle=1
870 #auth_param basic credentialsttl 2 hours
873 NAME: authenticate_cache_garbage_interval
877 LOC: Auth::TheConfig.garbageCollectInterval
879 The time period between garbage collection across the username cache.
880 This is a trade-off between memory utilization (long intervals - say
881 2 days) and CPU (short intervals - say 1 minute). Only change if you
885 NAME: authenticate_ttl
889 LOC: Auth::TheConfig.credentialsTtl
891 The time a user & their credentials stay in the logged in
892 user cache since their last request. When the garbage
893 interval passes, all user credentials that have passed their
894 TTL are removed from memory.
897 NAME: authenticate_ip_ttl
900 LOC: Auth::TheConfig.ipTtl
903 If you use proxy authentication and the 'max_user_ip' ACL,
904 this directive controls how long Squid remembers the IP
905 addresses associated with each user. Use a small value
906 (e.g., 60 seconds) if your users might change addresses
907 quickly, as is the case with dialup. You might be safe
908 using a larger value (e.g., 2 hours) in a corporate LAN
909 environment with relatively static address assignments.
914 -----------------------------------------------------------------------------
917 NAME: external_acl_type
918 TYPE: externalAclHelper
919 LOC: Config.externalAclHelperList
922 This option defines external acl classes using a helper program
923 to look up the status
925 external_acl_type name [options] FORMAT /path/to/helper [helper arguments]
929 ttl=n TTL in seconds for cached results (defaults to 3600
933 TTL for cached negative lookups (default same
936 grace=n Percentage remaining of TTL where a refresh of a
937 cached entry should be initiated without needing to
938 wait for a new reply. (default is for no grace period)
940 cache=n The maximum number of entries in the result cache. The
941 default limit is 262144 entries. Each cache entry usually
942 consumes at least 256 bytes. Squid currently does not remove
943 expired cache entries until the limit is reached, so a proxy
944 will sooner or later reach the limit. The expanded FORMAT
945 value is used as the cache key, so if the details in FORMAT
946 are highly variable, a larger cache may be needed to produce
947 reduction in helper load.
950 Maximum number of acl helper processes spawned to service
951 external acl lookups of this type. (default 5)
954 Minimum number of acl helper processes to spawn during
955 startup and reconfigure to service external acl lookups
956 of this type. (default 0)
959 Number of acl helper processes to keep ahead of traffic
960 loads. Squid will spawn this many at once whenever load
961 rises above the capabilities of existing processes.
962 Up to the value of children-max. (default 1)
964 concurrency=n concurrency level per process. Only used with helpers
965 capable of processing more than one query at a time.
967 queue-size=N The queue-size option sets the maximum number of
968 queued requests. A request is queued when no existing
969 helper can accept it due to concurrency limit and no
970 new helper can be started due to children-max limit.
971 If the queued requests exceed queue size, the acl is
972 ignored. The default value is set to 2*children-max.
974 protocol=2.5 Compatibility mode for Squid-2.5 external acl helpers.
976 ipv4 / ipv6 IP protocol used to communicate with this helper.
977 The default is to auto-detect IPv6 and use it when available.
980 FORMAT is a series of %macro codes. See logformat directive for a full list
981 of the accepted codes. Although note that at the time of any external ACL
982 being tested data may not be available and thus some %macro expand to '-'.
984 In addition to the logformat codes; when processing external ACLs these
985 additional macros are made available:
987 %ACL The name of the ACL being tested.
989 %DATA The ACL arguments specified in the referencing config
990 'acl ... external' line, separated by spaces (an
991 "argument string"). see acl external.
993 If there are no ACL arguments %DATA expands to '-'.
995 If you do not specify a DATA macro inside FORMAT,
996 Squid automatically appends %DATA to your FORMAT.
997 Note that Squid-3.x may expand %DATA to whitespace
998 or nothing in this case.
1000 By default, Squid applies URL-encoding to each ACL
1001 argument inside the argument string. If an explicit
1002 encoding modifier is used (e.g., %#DATA), then Squid
1003 encodes the whole argument string as a single token
1004 (e.g., with %#DATA, spaces between arguments become
1007 If SSL is enabled, the following formatting codes become available:
1009 %USER_CERT SSL User certificate in PEM format
1010 %USER_CERTCHAIN SSL User certificate chain in PEM format
1011 %USER_CERT_xx SSL User certificate subject attribute xx
1012 %USER_CA_CERT_xx SSL User certificate issuer attribute xx
1015 NOTE: all other format codes accepted by older Squid versions
1019 General request syntax:
1021 [channel-ID] FORMAT-values
1024 FORMAT-values consists of transaction details expanded with
1025 whitespace separation per the config file FORMAT specification
1026 using the FORMAT macros listed above.
1028 Request values sent to the helper are URL escaped to protect
1029 each value in requests against whitespaces.
1031 If using protocol=2.5 then the request sent to the helper is not
1032 URL escaped to protect against whitespace.
1034 NOTE: protocol=3.0 is deprecated as no longer necessary.
1036 When using the concurrency= option the protocol is changed by
1037 introducing a query channel tag in front of the request/response.
1038 The query channel tag is a number between 0 and concurrency-1.
1039 This value must be echoed back unchanged to Squid as the first part
1040 of the response relating to its request.
1043 The helper receives lines expanded per the above format specification
1044 and for each input line returns 1 line starting with OK/ERR/BH result
1045 code and optionally followed by additional keywords with more details.
1048 General result syntax:
1050 [channel-ID] result keyword=value ...
1052 Result consists of one of the codes:
1055 the ACL test produced a match.
1058 the ACL test does not produce a match.
1061 An internal error occurred in the helper, preventing
1062 a result being identified.
1064 The meaning of 'a match' is determined by your squid.conf
1065 access control configuration. See the Squid wiki for details.
1069 user= The users name (login)
1071 password= The users password (for login= cache_peer option)
1073 message= Message describing the reason for this response.
1074 Available as %o in error pages.
1075 Useful on (ERR and BH results).
1077 tag= Apply a tag to a request. Only sets a tag once,
1078 does not alter existing tags.
1080 log= String to be logged in access.log. Available as
1081 %ea in logformat specifications.
1083 clt_conn_tag= Associates a TAG with the client TCP connection.
1084 Please see url_rewrite_program related documentation
1087 Any keywords may be sent on any response whether OK, ERR or BH.
1089 All response keyword values need to be a single token with URL
1090 escaping, or enclosed in double quotes (") and escaped using \ on
1091 any double quotes or \ characters within the value. The wrapping
1092 double quotes are removed before the value is interpreted by Squid.
1093 \r and \n are also replace by CR and LF.
1095 Some example key values:
1099 user="J. \"Bob\" Smith"
1104 LOC: Config.namedAcls
1106 DEFAULT: ssl::certHasExpired ssl_error X509_V_ERR_CERT_HAS_EXPIRED
1107 DEFAULT: ssl::certNotYetValid ssl_error X509_V_ERR_CERT_NOT_YET_VALID
1108 DEFAULT: ssl::certDomainMismatch ssl_error SQUID_X509_V_ERR_DOMAIN_MISMATCH
1109 DEFAULT: ssl::certUntrusted ssl_error X509_V_ERR_INVALID_CA X509_V_ERR_SELF_SIGNED_CERT_IN_CHAIN X509_V_ERR_UNABLE_TO_VERIFY_LEAF_SIGNATURE X509_V_ERR_UNABLE_TO_GET_ISSUER_CERT X509_V_ERR_UNABLE_TO_GET_ISSUER_CERT_LOCALLY X509_V_ERR_CERT_UNTRUSTED
1110 DEFAULT: ssl::certSelfSigned ssl_error X509_V_ERR_DEPTH_ZERO_SELF_SIGNED_CERT
1112 DEFAULT: all src all
1113 DEFAULT: manager url_regex +i ^[^:]+://[^/]+/squid-internal-mgr/
1114 DEFAULT: localhost src 127.0.0.1/32 ::1
1115 DEFAULT: to_localhost dst 127.0.0.0/8 0.0.0.0/32 ::1/128 ::/128
1116 DEFAULT: to_linklocal dst 169.254.0.0/16 fe80::/10
1117 DEFAULT: CONNECT method CONNECT
1118 DEFAULT_DOC: ACLs all, manager, localhost, to_localhost, to_linklocal, and CONNECT are predefined.
1120 Defining an Access List
1122 Every access list definition must begin with an aclname and acltype,
1123 followed by either type-specific arguments or a quoted filename that
1126 acl aclname acltype argument ...
1127 acl aclname acltype "file" ...
1129 When using "file", the file should contain one item per line.
1134 Some acl types supports options which changes their default behaviour:
1136 -i,+i By default, regular expressions are CASE-SENSITIVE. To make them
1137 case-insensitive, use the -i option. To return case-sensitive
1138 use the +i option between patterns, or make a new ACL line
1141 -n Disable lookups and address type conversions. If lookup or
1142 conversion is required because the parameter type (IP or
1143 domain name) does not match the message address type (domain
1144 name or IP), then the ACL would immediately declare a mismatch
1145 without any warnings or lookups.
1148 Perform a list membership test, interpreting values as
1149 comma-separated token lists and matching against individual
1150 tokens instead of whole values.
1151 The optional "delimiters" parameter specifies one or more
1152 alternative non-alphanumeric delimiter characters.
1153 non-alphanumeric delimiter characters.
1155 -- Used to stop processing all options, in the case the first acl
1156 value has '-' character as first character (for example the '-'
1157 is a valid domain name)
1159 Some acl types require suspending the current request in order
1160 to access some external data source.
1161 Those which do are marked with the tag [slow], those which
1162 don't are marked as [fast].
1163 See https://wiki.squid-cache.org/SquidFaq/SquidAcl
1164 for further information
1166 ***** ACL TYPES AVAILABLE *****
1168 acl aclname src ip-address/mask ... # clients IP address [fast]
1169 acl aclname src addr1-addr2/mask ... # range of addresses [fast]
1170 acl aclname dst [-n] ip-address/mask ... # URL host's IP address [slow]
1171 acl aclname localip ip-address/mask ... # IP address the client connected to [fast]
1174 acl aclname arp mac-address ...
1175 acl aclname eui64 eui64-address ...
1177 # MAC (EUI-48) and EUI-64 addresses use xx:xx:xx:xx:xx:xx notation.
1179 # The 'arp' ACL code is not portable to all operating systems.
1180 # It works on Linux, Solaris, Windows, FreeBSD, and some other
1183 # The eui_lookup directive is required to be 'on' (the default)
1184 # and Squid built with --enable-eui for MAC/EUI addresses to be
1185 # available for this ACL.
1187 # Squid can only determine the MAC/EUI address for IPv4
1188 # clients that are on the same subnet. If the client is on a
1189 # different subnet, then Squid cannot find out its address.
1191 # IPv6 protocol does not contain ARP. MAC/EUI is either
1192 # encoded directly in the IPv6 address or not available.
1194 acl aclname clientside_mark mark[/mask] ...
1195 # matches CONNMARK of an accepted connection [fast]
1196 # DEPRECATED. Use the 'client_connection_mark' instead.
1198 acl aclname client_connection_mark mark[/mask] ...
1199 # matches CONNMARK of an accepted connection [fast]
1201 # mark and mask are unsigned integers (hex, octal, or decimal).
1202 # If multiple marks are given, then the ACL matches if at least
1205 # Uses netfilter-conntrack library.
1206 # Requires building Squid with --enable-linux-netfilter.
1208 # The client, various intermediaries, and Squid itself may set
1209 # CONNMARK at various times. The last CONNMARK set wins. This ACL
1210 # checks the mark present on an accepted connection or set by
1211 # Squid afterwards, depending on the ACL check timing. This ACL
1212 # effectively ignores any mark set by other agents after Squid has
1213 # accepted the connection.
1215 acl aclname srcdomain .foo.com ...
1216 # reverse lookup, from client IP [slow]
1217 acl aclname dstdomain [-n] .foo.com ...
1218 # Destination server from URL [fast]
1219 acl aclname srcdom_regex [-i] \.foo\.com ...
1220 # POSIX extended regex matching client name [slow]
1221 acl aclname dstdom_regex [-n] [-i] \.foo\.com ...
1222 # POSIX extended regex matching server [fast]
1224 # For dstdomain and dstdom_regex a reverse lookup is tried if a IP
1225 # based URL is used and no match is found. The name "none" is used
1226 # if the reverse lookup fails.
1228 acl aclname peername myPeer ...
1229 acl aclname peername_regex [-i] regex-pattern ...
1231 # match against a named cache_peer entry
1232 # set unique name= on cache_peer lines for reliable use.
1234 acl aclname time [day-abbrevs] [h1:m1-h2:m2]
1244 # h1:m1 must be less than h2:m2
1246 acl aclname url_regex [-i] ^http:// ...
1247 # POSIX extended regex matching on whole URL [fast]
1248 acl aclname urllogin [-i] [^a-zA-Z0-9] ...
1249 # POSIX extended regex matching on URL login field
1250 acl aclname urlpath_regex [-i] \.gif$ ...
1251 # POSIX extended regex matching on URL path [fast]
1253 acl aclname port 80 70 21 0-1024 ...
1254 # destination TCP port (or port range) of the request [fast]
1256 # Port 0 matches requests that have no explicit and no default destination
1257 # ports (e.g., HTTP requests with URN targets)
1259 acl aclname localport 3128 ... # TCP port the client connected to [fast]
1260 # NP: for interception mode this is usually '80'
1262 acl aclname myportname 3128 ... # *_port name [fast]
1264 acl aclname proto HTTP FTP ... # request protocol [fast]
1266 acl aclname method GET POST ... # HTTP request method [fast]
1268 acl aclname http_status 200 301 500- 400-403 ...
1269 # status code in reply [fast]
1271 acl aclname browser [-i] regex ...
1272 # POSIX extended regex match on User-Agent header
1273 # (see also req_header below) [fast]
1275 acl aclname referer_regex [-i] regex ...
1276 # POSIX extended regex match on Referer header [fast]
1277 # Referer is highly unreliable, so use with care
1279 acl aclname proxy_auth [-i] username ...
1280 # perform http authentication challenge to the client and match against
1281 # supplied credentials [slow]
1283 # takes a list of allowed usernames.
1284 # use REQUIRED to accept any valid username.
1286 # See proxy_auth_regex for more information. The two ACLs differ only in
1287 # their parameter syntax and username matching algorithm.
1289 acl aclname proxy_auth_regex [-i] username_pattern ...
1290 # perform http authentication challenge to the client and
1291 # POSIX extended regex match on supplied username [slow]
1293 # Will use proxy authentication in forward-proxy scenarios, and plain
1294 # http authentication in reverse-proxy scenarios
1296 # NOTE: when a Proxy-Authentication header is sent but it is not
1297 # needed during ACL checking the username is NOT logged
1300 # NOTE: proxy_auth requires a EXTERNAL authentication program
1301 # to check username/password combinations (see
1302 # auth_param directive).
1304 # NOTE: proxy_auth can't be used in a transparent/intercepting proxy
1305 # as the browser needs to be configured for using a proxy in order
1306 # to respond to proxy authentication.
1308 acl aclname snmp_community string ...
1309 # A community string to limit access to your SNMP Agent [fast]
1312 # acl snmppublic snmp_community public
1314 acl aclname maxconn number
1315 # This will be matched when the client's IP address has
1316 # more than <number> TCP connections established. [fast]
1317 # NOTE: This only measures direct TCP links so X-Forwarded-For
1318 # indirect clients are not counted.
1320 acl aclname max_user_ip [-s] number
1321 # This will be matched when the user attempts to log in from more
1322 # than <number> different ip addresses. The authenticate_ip_ttl
1323 # parameter controls the timeout on the ip entries. [fast]
1324 # If -s is specified the limit is strict, denying browsing
1325 # from any further IP addresses until the ttl has expired. Without
1326 # -s Squid will just annoy the user by "randomly" denying requests.
1327 # (the counter is reset each time the limit is reached and a
1328 # request is denied)
1329 # NOTE: in acceleration mode or where there is mesh of child proxies,
1330 # clients may appear to come from multiple addresses if they are
1331 # going through proxy farms, so a limit of 1 may cause user problems.
1333 acl aclname random probability
1334 # Pseudo-randomly match requests. Based on the probability given.
1335 # Probability may be written as a decimal (0.333), fraction (1/3)
1336 # or ratio of matches:non-matches (3:5).
1338 acl aclname req_mime_type [-i] mime-type ...
1339 # POSIX extended regex match against the mime type of the request generated
1340 # by the client. Can be used to detect file upload or some
1341 # types HTTP tunneling requests [fast]
1342 # NOTE: This does NOT match the reply. You cannot use this
1343 # to match the returned file type.
1345 acl aclname req_header header-name [-i] any\.regex\.here
1346 # POSIX extended regex match against any of the known request headers. May be
1347 # thought of as a superset of "browser", "referer" and "mime-type"
1350 acl aclname rep_mime_type [-i] mime-type ...
1351 # POSIX extended regex match against the mime type of the reply received by
1352 # squid. Can be used to detect file download or some
1353 # types HTTP tunneling requests. [fast]
1354 # NOTE: This has no effect in http_access rules. It only has
1355 # effect in rules that affect the reply data stream such as
1356 # http_reply_access.
1358 acl aclname rep_header header-name [-i] any\.regex\.here
1359 # POSIX extended regex match against any of the known reply headers. May be
1360 # thought of as a superset of "browser", "referer" and "mime-type"
1363 acl aclname external class_name [arguments...]
1364 # external ACL lookup via a helper class defined by the
1365 # external_acl_type directive [slow]
1367 acl aclname user_cert attribute values...
1368 # match against attributes in a user SSL certificate
1369 # attribute is one of DN/C/O/CN/L/ST or a numerical OID [fast]
1371 acl aclname ca_cert attribute values...
1372 # match against attributes a users issuing CA SSL certificate
1373 # attribute is one of DN/C/O/CN/L/ST or a numerical OID [fast]
1375 acl aclname ext_user [-i] username ...
1376 # string match on username returned by external acl helper [slow]
1377 # use REQUIRED to accept any non-null user name.
1379 # See also: ext_user_regex. The two ACLs differ only in their parameter
1380 # syntax and username matching algorithm.
1382 acl aclname ext_user_regex [-i] username_pattern ...
1383 # POSIX extended regex match on username returned by external acl helper [slow]
1385 acl aclname tag tagvalue ...
1386 # string match on tag returned by external acl helper [fast]
1387 # DEPRECATED. Only the first tag will match with this ACL.
1388 # Use the 'note' ACL instead for handling multiple tag values.
1390 acl aclname hier_code codename ...
1391 # string match against squid hierarchy code(s); [fast]
1392 # e.g., DIRECT, PARENT_HIT, NONE, etc.
1394 # NOTE: This has no effect in http_access rules. It only has
1395 # effect in rules that affect the reply data stream such as
1396 # http_reply_access.
1398 acl aclname note [-m[=delimiters]] name [value ...]
1399 # match transaction annotation [fast]
1400 # Without values, matches any annotation with a given name.
1401 # With value(s), matches any annotation with a given name that
1402 # also has one of the given values.
1403 # If the -m flag is used, then the value of the named
1404 # annotation is interpreted as a list of tokens, and the ACL
1405 # matches individual name=token pairs rather than whole
1406 # name=value pairs. See "ACL Options" above for more info.
1407 # Annotation sources include note and adaptation_meta directives
1408 # as well as helper and eCAP responses.
1410 acl aclname annotate_transaction [-m[=delimiters]] key=value ...
1411 acl aclname annotate_transaction [-m[=delimiters]] key+=value ...
1412 # Always matches. [fast]
1413 # Used for its side effect: This ACL immediately adds a
1414 # key=value annotation to the current master transaction.
1415 # The added annotation can then be tested using note ACL and
1416 # logged (or sent to helpers) using %note format code.
1418 # Annotations can be specified using replacement and addition
1419 # formats. The key=value form replaces old same-key annotation
1420 # value(s). The key+=value form appends a new value to the old
1421 # same-key annotation. Both forms create a new key=value
1422 # annotation if no same-key annotation exists already. If
1423 # -m flag is used, then the value is interpreted as a list
1424 # and the annotation will contain key=token pair(s) instead of the
1425 # whole key=value pair.
1427 # This ACL is especially useful for recording complex multi-step
1428 # ACL-driven decisions. For example, the following configuration
1429 # avoids logging transactions accepted after aclX matched:
1431 # # First, mark transactions accepted after aclX matched
1432 # acl markSpecial annotate_transaction special=true
1433 # http_access allow acl001
1435 # http_access deny acl100
1436 # http_access allow aclX markSpecial
1438 # # Second, do not log marked transactions:
1439 # acl markedSpecial note special true
1440 # access_log ... deny markedSpecial
1442 # # Note that the following would not have worked because aclX
1443 # # alone does not determine whether the transaction was allowed:
1444 # access_log ... deny aclX # Wrong!
1446 # Warning: This ACL annotates the transaction even when negated
1447 # and even if subsequent ACLs fail to match. For example, the
1448 # following three rules will have exactly the same effect as far
1449 # as annotations set by the "mark" ACL are concerned:
1451 # some_directive acl1 ... mark # rule matches if mark is reached
1452 # some_directive acl1 ... !mark # rule never matches
1453 # some_directive acl1 ... mark !all # rule never matches
1455 acl aclname annotate_client [-m[=delimiters]] key=value ...
1456 acl aclname annotate_client [-m[=delimiters]] key+=value ...
1458 # Always matches. [fast]
1459 # Used for its side effect: This ACL immediately adds a
1460 # key=value annotation to the current client-to-Squid
1461 # connection. Connection annotations are propagated to the current
1462 # and all future master transactions on the annotated connection.
1463 # See the annotate_transaction ACL for details.
1465 # For example, the following configuration avoids rewriting URLs
1466 # of transactions bumped by SslBump:
1468 # # First, mark bumped connections:
1469 # acl markBumped annotate_client bumped=true
1470 # ssl_bump peek acl1
1471 # ssl_bump stare acl2
1472 # ssl_bump bump acl3 markBumped
1473 # ssl_bump splice all
1475 # # Second, do not send marked transactions to the redirector:
1476 # acl markedBumped note bumped true
1477 # url_rewrite_access deny markedBumped
1479 # # Note that the following would not have worked because acl3 alone
1480 # # does not determine whether the connection is going to be bumped:
1481 # url_rewrite_access deny acl3 # Wrong!
1483 acl aclname adaptation_service service ...
1484 # Matches the name of any icap_service, ecap_service,
1485 # adaptation_service_set, or adaptation_service_chain that Squid
1486 # has used (or attempted to use) for the master transaction.
1487 # This ACL must be defined after the corresponding adaptation
1488 # service is named in squid.conf. This ACL is usable with
1489 # adaptation_meta because it starts matching immediately after
1490 # the service has been selected for adaptation.
1492 acl aclname transaction_initiator initiator ...
1493 # Matches transaction's initiator [fast]
1495 # Supported initiators are:
1496 # certificate-fetching: matches transactions fetching
1497 # a missing intermediate TLS certificate
1498 # cache-digest: matches transactions fetching Cache Digests
1500 # htcp: matches HTCP requests from peers
1501 # icp: matches ICP requests to peers
1502 # icmp: matches ICMP RTT database (NetDB) requests to peers
1503 # internal: matches any of the above
1504 # client: matches transactions containing an HTTP or FTP
1505 # client request received at a Squid *_port
1506 # all: matches any transaction, including internal transactions
1507 # without a configurable initiator and hopefully rare
1508 # transactions without a known-to-Squid initiator
1510 # Multiple initiators are ORed.
1512 acl aclname has component
1513 # matches a transaction "component" [fast]
1515 # Supported transaction components are:
1516 # request: transaction has a request header (at least)
1517 # response: transaction has a response header (at least)
1518 # ALE: transaction has an internally-generated Access Log Entry
1519 # structure; bugs notwithstanding, all transaction have it
1521 # For example, the following configuration helps when dealing with HTTP
1522 # clients that close connections without sending a request header:
1524 # acl hasRequest has request
1525 # acl logMe note important_transaction
1526 # # avoid "logMe ACL is used in context without an HTTP request" warnings
1527 # access_log ... logformat=detailed hasRequest logMe
1528 # # log request-less transactions, instead of ignoring them
1529 # access_log ... logformat=brief !hasRequest
1531 # Multiple components are not supported for one "acl" rule, but
1532 # can be specified (and are ORed) using multiple same-name rules:
1534 # # OK, this strange logging daemon needs request or response,
1535 # # but can work without either a request or a response:
1536 # acl hasWhatMyLoggingDaemonNeeds has request
1537 # acl hasWhatMyLoggingDaemonNeeds has response
1539 acl aclname at_step step
1540 # match against the current request processing step [fast]
1542 # GeneratingCONNECT: Generating HTTP CONNECT request headers
1544 # The following ssl_bump processing steps are recognized:
1545 # SslBump1: After getting TCP-level and HTTP CONNECT info.
1546 # SslBump2: After getting SSL Client Hello info.
1547 # SslBump3: After getting SSL Server Hello info.
1551 acl aclname ssl_error errorname
1552 # match against SSL certificate validation error [fast]
1554 # When used with sslproxy_cert_error, this ACL tests a single
1555 # certificate validation error currently being evaluated by that
1556 # directive. When used with slproxy_cert_sign or sslproxy_cert_adapt,
1557 # the ACL tests all past certificate validation errors associated with
1558 # the current Squid-to-server connection (attempt). This ACL is not yet
1559 # supported for use with other directives.
1561 # For valid error names see in @DEFAULT_ERROR_DIR@/templates/error-details.txt
1564 # The following can be used as shortcuts for certificate properties:
1565 # [ssl::]certHasExpired: the "not after" field is in the past
1566 # [ssl::]certNotYetValid: the "not before" field is in the future
1567 # [ssl::]certUntrusted: The certificate issuer is not to be trusted.
1568 # [ssl::]certSelfSigned: The certificate is self signed.
1569 # [ssl::]certDomainMismatch: The certificate CN domain does not
1570 # match the name the name of the host we are connecting to.
1572 # The ssl::certHasExpired, ssl::certNotYetValid, ssl::certDomainMismatch,
1573 # ssl::certUntrusted, and ssl::certSelfSigned can also be used as
1574 # predefined ACLs, just like the 'all' ACL.
1576 acl aclname server_cert_fingerprint fingerprint
1577 # match against server SSL certificate fingerprint [fast]
1579 # The fingerprint is the digest of the DER encoded version
1580 # of the whole certificate. The user should use the form: XX:XX:...
1581 # The SHA1 digest algorithm is the default and is currently
1582 # the only algorithm supported.
1584 acl aclname ssl::server_name [option] .foo.com ...
1585 # matches server name obtained from various sources [fast]
1587 # The ACL computes server name(s) using such information sources as
1588 # CONNECT request URI, TLS client SNI, and TLS server certificate
1589 # subject (CN and SubjectAltName). The computed server name(s) usually
1590 # change with each SslBump step, as more info becomes available:
1591 # * SNI is used as the server name instead of the request URI,
1592 # * subject name(s) from the server certificate (CN and
1593 # SubjectAltName) are used as the server names instead of SNI.
1595 # When the ACL computes multiple server names, matching any single
1596 # computed name is sufficient for the ACL to match.
1598 # The "none" name can be used to match transactions where the ACL
1599 # could not compute the server name using any information source
1600 # that was both available and allowed to be used by the ACL options at
1601 # the ACL evaluation time.
1603 # Unlike dstdomain, this ACL does not perform DNS lookups.
1605 # A server name may be an IP address. For example, subject alternative
1606 # names (a.k.a. SANs) in some real server certificates include IPv4 and
1607 # IPv6 entries. Internally, Squid uses inet_ntop(3) to prep IP names for
1608 # matching. When using IPv6 names, use "::" notation (if applicable).
1609 # Do not use brackets. For example: 1080::8:800:200c:417a.
1611 # An ACL option below may be used to restrict what information
1612 # sources are used to extract the server names from:
1614 # --client-requested
1615 # The server name is SNI regardless of what the server says.
1617 # The server name(s) are the certificate subject name(s), regardless
1618 # of what the client has requested. If the server certificate is
1619 # unavailable, then the name is "none".
1621 # The server name is either SNI (if SNI matches at least one of the
1622 # certificate subject names) or "none" (otherwise). When the server
1623 # certificate is unavailable, the consensus server name is SNI.
1625 # Combining multiple options in one ACL is a fatal configuration
1628 # For all options: If no SNI is available, then the CONNECT request
1629 # target (a.k.a. URI) is used instead of SNI (for an intercepted
1630 # connection, this target is the destination IP address).
1632 acl aclname ssl::server_name_regex [-i] \.foo\.com ...
1633 # POSIX extended regex matches server name obtained from various sources [fast]
1635 # See ssl::server_name for details, including IPv6 address formatting
1636 # caveats. Use case-insensitive matching (i.e. -i option) to reduce
1637 # dependency on how Squid formats or sanitizes server names.
1639 acl aclname connections_encrypted
1640 # matches transactions with all HTTP messages received over TLS
1641 # transport connections. [fast]
1643 # The master transaction deals with HTTP messages received from
1644 # various sources. All sources used by the master transaction in the
1645 # past are considered by the ACL. The following rules define whether
1646 # a given message source taints the entire master transaction,
1647 # resulting in ACL mismatches:
1649 # * The HTTP client transport connection is not TLS.
1650 # * An adaptation service connection-encryption flag is off.
1651 # * The peer or origin server transport connection is not TLS.
1653 # Caching currently does not affect these rules. This cache ignorance
1654 # implies that only the current HTTP client transport and REQMOD
1655 # services status determine whether this ACL matches a from-cache
1656 # transaction. The source of the cached response does not have any
1657 # effect on future transaction that use the cached response without
1658 # revalidation. This may change.
1660 # DNS, ICP, and HTCP exchanges during the master transaction do not
1661 # affect these rules.
1663 acl aclname any-of acl1 acl2 ...
1664 # match any one of the acls [fast or slow]
1665 # The first matching ACL stops further ACL evaluation.
1667 # ACLs from multiple any-of lines with the same name are ORed.
1668 # For example, A = (a1 or a2) or (a3 or a4) can be written as
1669 # acl A any-of a1 a2
1670 # acl A any-of a3 a4
1672 # This group ACL is fast if all evaluated ACLs in the group are fast
1673 # and slow otherwise.
1675 acl aclname all-of acl1 acl2 ...
1676 # match all of the acls [fast or slow]
1677 # The first mismatching ACL stops further ACL evaluation.
1679 # ACLs from multiple all-of lines with the same name are ORed.
1680 # For example, B = (b1 and b2) or (b3 and b4) can be written as
1681 # acl B all-of b1 b2
1682 # acl B all-of b3 b4
1684 # This group ACL is fast if all evaluated ACLs in the group are fast
1685 # and slow otherwise.
1688 acl macaddress arp 09:00:2b:23:45:67
1689 acl password proxy_auth REQUIRED
1690 acl fileupload req_mime_type -i ^multipart/form-data$
1691 acl javascript rep_mime_type -i ^application/x-javascript$
1695 # Recommended minimum configuration:
1698 # Example rule allowing access from your local networks.
1699 # Adapt to list your (internal) IP networks from where browsing
1701 acl localnet src 0.0.0.1-0.255.255.255 # RFC 1122 "this" network (LAN)
1702 acl localnet src 10.0.0.0/8 # RFC 1918 local private network (LAN)
1703 acl localnet src 100.64.0.0/10 # RFC 6598 shared address space (CGN)
1704 acl localnet src 169.254.0.0/16 # RFC 3927 link-local (directly plugged) machines
1705 acl localnet src 172.16.0.0/12 # RFC 1918 local private network (LAN)
1706 acl localnet src 192.168.0.0/16 # RFC 1918 local private network (LAN)
1707 acl localnet src fc00::/7 # RFC 4193 local private network range
1708 acl localnet src fe80::/10 # RFC 4291 link-local (directly plugged) machines
1710 acl SSL_ports port 443
1711 acl Safe_ports port 80 # http
1712 acl Safe_ports port 21 # ftp
1713 acl Safe_ports port 443 # https
1714 acl Safe_ports port 70 # gopher
1715 acl Safe_ports port 210 # wais
1716 acl Safe_ports port 1025-65535 # unregistered ports
1717 acl Safe_ports port 280 # http-mgmt
1718 acl Safe_ports port 488 # gss-http
1719 acl Safe_ports port 591 # filemaker
1720 acl Safe_ports port 777 # multiling http
1724 NAME: proxy_protocol_access
1726 LOC: Config.accessList.proxyProtocol
1728 DEFAULT_DOC: all TCP connections to ports with require-proxy-header will be denied
1730 Determine which client proxies can be trusted to provide correct
1731 information regarding real client IP address using PROXY protocol.
1733 Requests may pass through a chain of several other proxies
1734 before reaching us. The original source details may by sent in:
1735 * HTTP message Forwarded header, or
1736 * HTTP message X-Forwarded-For header, or
1737 * PROXY protocol connection header.
1739 This directive is solely for validating new PROXY protocol
1740 connections received from a port flagged with require-proxy-header.
1741 It is checked only once after TCP connection setup.
1743 A deny match results in TCP connection closure.
1745 An allow match is required for Squid to permit the corresponding
1746 TCP connection, before Squid even looks for HTTP request headers.
1747 If there is an allow match, Squid starts using PROXY header information
1748 to determine the source address of the connection for all future ACL
1749 checks, logging, etc.
1751 SECURITY CONSIDERATIONS:
1753 Any host from which we accept client IP details can place
1754 incorrect information in the relevant header, and Squid
1755 will use the incorrect information as if it were the
1756 source address of the request. This may enable remote
1757 hosts to bypass any access control restrictions that are
1758 based on the client's source addresses.
1760 This clause only supports fast acl types.
1761 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
1764 NAME: follow_x_forwarded_for
1766 IFDEF: FOLLOW_X_FORWARDED_FOR
1767 LOC: Config.accessList.followXFF
1768 DEFAULT_IF_NONE: deny all
1769 DEFAULT_DOC: X-Forwarded-For header will be ignored.
1771 Determine which client proxies can be trusted to provide correct
1772 information regarding real client IP address.
1774 Requests may pass through a chain of several other proxies
1775 before reaching us. The original source details may by sent in:
1776 * HTTP message Forwarded header, or
1777 * HTTP message X-Forwarded-For header, or
1778 * PROXY protocol connection header.
1780 PROXY protocol connections are controlled by the proxy_protocol_access
1781 directive which is checked before this.
1783 If a request reaches us from a source that is allowed by this
1784 directive, then we trust the information it provides regarding
1785 the IP of the client it received from (if any).
1787 For the purpose of ACLs used in this directive the src ACL type always
1788 matches the address we are testing and srcdomain matches its rDNS.
1790 On each HTTP request Squid checks for X-Forwarded-For header fields.
1791 If found the header values are iterated in reverse order and an allow
1792 match is required for Squid to continue on to the next value.
1793 The verification ends when a value receives a deny match, cannot be
1794 tested, or there are no more values to test.
1795 NOTE: Squid does not yet follow the Forwarded HTTP header.
1797 The end result of this process is an IP address that we will
1798 refer to as the indirect client address. This address may
1799 be treated as the client address for access control, ICAP, delay
1800 pools and logging, depending on the acl_uses_indirect_client,
1801 icap_uses_indirect_client, delay_pool_uses_indirect_client,
1802 log_uses_indirect_client and tproxy_uses_indirect_client options.
1804 This clause only supports fast acl types.
1805 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
1807 SECURITY CONSIDERATIONS:
1809 Any host from which we accept client IP details can place
1810 incorrect information in the relevant header, and Squid
1811 will use the incorrect information as if it were the
1812 source address of the request. This may enable remote
1813 hosts to bypass any access control restrictions that are
1814 based on the client's source addresses.
1818 acl localhost src 127.0.0.1
1819 acl my_other_proxy srcdomain .proxy.example.com
1820 follow_x_forwarded_for allow localhost
1821 follow_x_forwarded_for allow my_other_proxy
1824 NAME: acl_uses_indirect_client
1827 IFDEF: FOLLOW_X_FORWARDED_FOR
1829 LOC: Config.onoff.acl_uses_indirect_client
1831 Controls whether the indirect client address
1832 (see follow_x_forwarded_for) is used instead of the
1833 direct client address in acl matching.
1835 NOTE: maxconn ACL considers direct TCP links and indirect
1836 clients will always have zero. So no match.
1839 NAME: delay_pool_uses_indirect_client
1842 IFDEF: FOLLOW_X_FORWARDED_FOR&&USE_DELAY_POOLS
1844 LOC: Config.onoff.delay_pool_uses_indirect_client
1846 Controls whether the indirect client address
1847 (see follow_x_forwarded_for) is used instead of the
1848 direct client address in delay pools.
1851 NAME: log_uses_indirect_client
1854 IFDEF: FOLLOW_X_FORWARDED_FOR
1856 LOC: Config.onoff.log_uses_indirect_client
1858 Controls whether the indirect client address
1859 (see follow_x_forwarded_for) is used instead of the
1860 direct client address in the access log.
1863 NAME: tproxy_uses_indirect_client
1866 IFDEF: FOLLOW_X_FORWARDED_FOR&&LINUX_NETFILTER
1868 LOC: Config.onoff.tproxy_uses_indirect_client
1870 Controls whether the indirect client address
1871 (see follow_x_forwarded_for) is used instead of the
1872 direct client address when spoofing the outgoing client.
1874 This has no effect on requests arriving in non-tproxy
1877 SECURITY WARNING: Usage of this option is dangerous
1878 and should not be used trivially. Correct configuration
1879 of follow_x_forwarded_for with a limited set of trusted
1880 sources is required to prevent abuse of your proxy.
1883 NAME: spoof_client_ip
1885 LOC: Config.accessList.spoof_client_ip
1887 DEFAULT_DOC: Allow spoofing on all TPROXY traffic.
1889 Control client IP address spoofing of TPROXY traffic based on
1890 defined access lists.
1892 spoof_client_ip allow|deny [!]aclname ...
1894 If there are no "spoof_client_ip" lines present, the default
1895 is to "allow" spoofing of any suitable request.
1897 Note that the cache_peer "no-tproxy" option overrides this ACL.
1899 This clause supports fast acl types.
1900 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
1905 LOC: Config.accessList.http
1906 DEFAULT_IF_NONE: deny all
1907 DEFAULT_DOC: Deny, unless rules exist in squid.conf.
1909 Allowing or Denying access based on defined access lists
1911 To allow or deny a message received on an HTTP, HTTPS, or FTP port:
1912 http_access allow|deny [!]aclname ...
1914 NOTE on default values:
1916 If there are no "access" lines present, the default is to deny
1919 If none of the "access" lines cause a match, the default is the
1920 opposite of the last line in the list. If the last line was
1921 deny, the default is allow. Conversely, if the last line
1922 is allow, the default will be deny. For these reasons, it is a
1923 good idea to have an "deny all" entry at the end of your access
1924 lists to avoid potential confusion.
1926 This clause supports both fast and slow acl types.
1927 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
1932 # Recommended minimum Access Permission configuration:
1934 # Deny requests to certain unsafe ports
1935 http_access deny !Safe_ports
1937 # Deny CONNECT to other than secure SSL ports
1938 http_access deny CONNECT !SSL_ports
1940 # Only allow cachemgr access from localhost
1941 http_access allow localhost manager
1942 http_access deny manager
1944 # This default configuration only allows localhost requests because a more
1945 # permissive Squid installation could introduce new attack vectors into the
1946 # network by proxying external TCP connections to unprotected services.
1947 http_access allow localhost
1949 # The two deny rules below are unnecessary in this default configuration
1950 # because they are followed by a "deny all" rule. However, they may become
1951 # critically important when you start allowing external requests below them.
1953 # Protect web applications running on the same server as Squid. They often
1954 # assume that only local users can access them at "localhost" ports.
1955 http_access deny to_localhost
1957 # Protect cloud servers that provide local users with sensitive info about
1958 # their server via certain well-known link-local (a.k.a. APIPA) addresses.
1959 http_access deny to_linklocal
1962 # INSERT YOUR OWN RULE(S) HERE TO ALLOW ACCESS FROM YOUR CLIENTS
1965 # For example, to allow access from your local networks, you may uncomment the
1966 # following rule (and/or add rules that match your definition of "local"):
1967 # http_access allow localnet
1969 # And finally deny all other access to this proxy
1970 http_access deny all
1974 NAME: adapted_http_access http_access2
1976 LOC: Config.accessList.adapted_http
1978 DEFAULT_DOC: Allow, unless rules exist in squid.conf.
1980 Allowing or Denying access based on defined access lists
1982 Essentially identical to http_access, but runs after redirectors
1983 and ICAP/eCAP adaptation. Allowing access control based on their
1986 If not set then only http_access is used.
1989 NAME: http_reply_access
1991 LOC: Config.accessList.reply
1993 DEFAULT_DOC: Allow, unless rules exist in squid.conf.
1995 Allow replies to client requests. This is complementary to http_access.
1997 http_reply_access allow|deny [!] aclname ...
1999 NOTE: if there are no access lines present, the default is to allow
2002 If none of the access lines cause a match the opposite of the
2003 last line will apply. Thus it is good practice to end the rules
2004 with an "allow all" or "deny all" entry.
2006 This clause supports both fast and slow acl types.
2007 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
2012 LOC: Config.accessList.icp
2014 DEFAULT_DOC: Deny, unless rules exist in squid.conf.
2016 Allowing or Denying access to the ICP port based on defined
2019 icp_access allow|deny [!]aclname ...
2021 NOTE: The default if no icp_access lines are present is to
2022 deny all traffic. This default may cause problems with peers
2025 This clause only supports fast acl types.
2026 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
2028 # Allow ICP queries from local networks only
2029 #icp_access allow localnet
2030 #icp_access deny all
2036 LOC: Config.accessList.htcp
2038 DEFAULT_DOC: Deny, unless rules exist in squid.conf.
2040 Allowing or Denying access to the HTCP port based on defined
2043 htcp_access allow|deny [!]aclname ...
2045 See also htcp_clr_access for details on access control for
2046 cache purge (CLR) HTCP messages.
2048 NOTE: The default if no htcp_access lines are present is to
2049 deny all traffic. This default may cause problems with peers
2050 using the htcp option.
2052 This clause only supports fast acl types.
2053 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
2055 # Allow HTCP queries from local networks only
2056 #htcp_access allow localnet
2057 #htcp_access deny all
2060 NAME: htcp_clr_access
2063 LOC: Config.accessList.htcp_clr
2065 DEFAULT_DOC: Deny, unless rules exist in squid.conf.
2067 Allowing or Denying access to purge content using HTCP based
2068 on defined access lists.
2069 See htcp_access for details on general HTCP access control.
2071 htcp_clr_access allow|deny [!]aclname ...
2073 This clause only supports fast acl types.
2074 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
2076 # Allow HTCP CLR requests from trusted peers
2077 acl htcp_clr_peer src 192.0.2.2 2001:DB8::2
2078 htcp_clr_access allow htcp_clr_peer
2079 htcp_clr_access deny all
2084 LOC: Config.accessList.miss
2086 DEFAULT_DOC: Allow, unless rules exist in squid.conf.
2088 Determines whether network access is permitted when satisfying a request.
2091 to force your neighbors to use you as a sibling instead of
2094 acl localclients src 192.0.2.0/24 2001:DB8::a:0/64
2095 miss_access deny !localclients
2096 miss_access allow all
2098 This means only your local clients are allowed to fetch relayed/MISS
2099 replies from the network and all other clients can only fetch cached
2102 The default for this setting allows all clients who passed the
2103 http_access rules to relay via this proxy.
2105 This clause only supports fast acl types.
2106 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
2109 NAME: reply_body_max_size
2110 COMMENT: size [acl acl...]
2113 DEFAULT_DOC: No limit is applied.
2114 LOC: Config.ReplyBodySize
2116 This option specifies the maximum size of a reply body. It can be
2117 used to prevent users from downloading very large files, such as
2118 MP3's and movies. When the reply headers are received, the
2119 reply_body_max_size lines are processed, and the first line where
2120 all (if any) listed ACLs are true is used as the maximum body size
2123 This size is checked twice. First when we get the reply headers,
2124 we check the content-length value. If the content length value exists
2125 and is larger than the allowed size, the request is denied and the
2126 user receives an error message that says "the request or reply
2127 is too large." If there is no content-length, and the reply
2128 size exceeds this limit, the client's connection is just closed
2129 and they will receive a partial reply.
2131 WARNING: downstream caches probably can not detect a partial reply
2132 if there is no content-length header, so they will cache
2133 partial responses and give them out as hits. You should NOT
2134 use this option if you have downstream caches.
2136 WARNING: A maximum size smaller than the size of squid's error messages
2137 will cause an infinite loop and crash squid. Ensure that the smallest
2138 non-zero value you use is greater that the maximum header size plus
2139 the size of your largest error page.
2141 If you set this parameter none (the default), there will be
2144 Configuration Format is:
2145 reply_body_max_size SIZE UNITS [acl ...]
2147 reply_body_max_size 10 MB
2151 NAME: on_unsupported_protocol
2152 TYPE: on_unsupported_protocol
2153 LOC: Config.accessList.on_unsupported_protocol
2155 DEFAULT_DOC: Respond with an error message to unidentifiable traffic
2157 Determines Squid behavior when encountering strange requests at the
2158 beginning of an accepted TCP connection or the beginning of a bumped
2159 CONNECT tunnel. Controlling Squid reaction to unexpected traffic is
2160 especially useful in interception environments where Squid is likely
2161 to see connections for unsupported protocols that Squid should either
2162 terminate or tunnel at TCP level.
2164 on_unsupported_protocol <action> [!]acl ...
2166 The first matching action wins. Only fast ACLs are supported.
2168 Supported actions are:
2170 tunnel: Establish a TCP connection with the intended server and
2171 blindly shovel TCP packets between the client and server.
2173 respond: Respond with an error message, using the transfer protocol
2174 for the Squid port that received the request (e.g., HTTP
2175 for connections intercepted at the http_port). This is the
2178 Squid expects the following traffic patterns:
2180 http_port: a plain HTTP request
2181 https_port: SSL/TLS handshake followed by an [encrypted] HTTP request
2182 ftp_port: a plain FTP command (no on_unsupported_protocol support yet!)
2183 CONNECT tunnel on http_port: same as https_port
2184 CONNECT tunnel on https_port: same as https_port
2186 Currently, this directive has effect on intercepted connections and
2187 bumped tunnels only. Other cases are not supported because Squid
2188 cannot know the intended destination of other traffic.
2191 # define what Squid errors indicate receiving non-HTTP traffic:
2192 acl foreignProtocol squid_error ERR_PROTOCOL_UNKNOWN ERR_TOO_BIG
2193 # define what Squid errors indicate receiving nothing:
2194 acl serverTalksFirstProtocol squid_error ERR_REQUEST_START_TIMEOUT
2195 # tunnel everything that does not look like HTTP:
2196 on_unsupported_protocol tunnel foreignProtocol
2197 # tunnel if we think the client waits for the server to talk first:
2198 on_unsupported_protocol tunnel serverTalksFirstProtocol
2199 # in all other error cases, just send an HTTP "error page" response:
2200 on_unsupported_protocol respond all
2202 See also: squid_error ACL
2208 LOC: Auth::TheConfig.schemeAccess
2210 DEFAULT_DOC: use all auth_param schemes in their configuration order
2212 Use this directive to customize authentication schemes presence and
2213 order in Squid's Unauthorized and Authentication Required responses.
2215 auth_schemes scheme1,scheme2,... [!]aclname ...
2217 where schemeN is the name of one of the authentication schemes
2218 configured using auth_param directives. At least one scheme name is
2219 required. Multiple scheme names are separated by commas. Either
2220 avoid whitespace or quote the entire schemes list.
2222 A special "ALL" scheme name expands to all auth_param-configured
2223 schemes in their configuration order. This directive cannot be used
2224 to configure Squid to offer no authentication schemes at all.
2226 The first matching auth_schemes rule determines the schemes order
2227 for the current Authentication Required transaction. Note that the
2228 future response is not yet available during auth_schemes evaluation.
2230 If this directive is not used or none of its rules match, then Squid
2231 responds with all configured authentication schemes in the order of
2232 auth_param directives in the configuration file.
2234 This directive does not determine when authentication is used or
2235 how each authentication scheme authenticates clients.
2237 The following example sends basic and negotiate authentication
2238 schemes, in that order, when requesting authentication of HTTP
2239 requests matching the isIE ACL (not shown) while sending all
2240 auth_param schemes in their configuration order to other clients:
2242 auth_schemes basic,negotiate isIE
2243 auth_schemes ALL all # explicit default
2245 This directive supports fast ACLs only.
2247 See also: auth_param.
2252 -----------------------------------------------------------------------------
2255 NAME: http_port ascii_port
2260 Usage: port [mode] [options]
2261 hostname:port [mode] [options]
2262 1.2.3.4:port [mode] [options]
2264 The socket addresses where Squid will listen for HTTP client
2265 requests. You may specify multiple socket addresses.
2266 There are three forms: port alone, hostname with port, and
2267 IP address with port. If you specify a hostname or IP
2268 address, Squid binds the socket to that specific
2269 address. Most likely, you do not need to bind to a specific
2270 address, so you can use the port number alone.
2272 If you are running Squid in accelerator mode, you
2273 probably want to listen on port 80 also, or instead.
2275 The -a command line option may be used to specify additional
2276 port(s) where Squid listens for proxy request. Such ports will
2277 be plain proxy ports with no options.
2279 You may specify multiple socket addresses on multiple lines.
2283 intercept Support for IP-Layer NAT interception delivering
2284 traffic to this Squid port.
2285 NP: disables authentication on the port.
2287 tproxy Support Linux TPROXY (or BSD divert-to) with spoofing
2288 of outgoing connections using the client IP address.
2289 NP: disables authentication on the port.
2291 accel Accelerator / reverse proxy mode
2293 ssl-bump For each CONNECT request allowed by ssl_bump ACLs,
2294 establish secure connection with the client and with
2295 the server, decrypt HTTPS messages as they pass through
2296 Squid, and treat them as unencrypted HTTP messages,
2297 becoming the man-in-the-middle.
2299 The ssl_bump option is required to fully enable
2300 bumping of CONNECT requests.
2302 Omitting the mode flag causes default forward proxy mode to be used.
2305 Accelerator Mode Options:
2307 defaultsite=domainname
2308 What to use for the Host: header if it is not present
2309 in a request. Determines what site (not origin server)
2310 accelerators should consider the default.
2312 no-vhost Disable using HTTP/1.1 Host header for virtual domain support.
2314 protocol= Protocol to reconstruct accelerated and intercepted
2315 requests with. Defaults to HTTP/1.1 for http_port and
2316 HTTPS/1.1 for https_port.
2317 When an unsupported value is configured Squid will
2318 produce a FATAL error.
2319 Values: HTTP or HTTP/1.1, HTTPS or HTTPS/1.1
2321 vport Virtual host port support. Using the http_port number
2322 instead of the port passed on Host: headers.
2324 vport=NN Virtual host port support. Using the specified port
2325 number instead of the port passed on Host: headers.
2328 Act as if this Squid is the origin server.
2329 This currently means generate new Date: and Expires:
2330 headers on HIT instead of adding Age:.
2332 ignore-cc Ignore request Cache-Control headers.
2334 WARNING: This option violates HTTP specifications if
2335 used in non-accelerator setups.
2337 allow-direct Allow direct forwarding in accelerator mode. Normally
2338 accelerated requests are denied direct forwarding as if
2339 never_direct was used.
2341 WARNING: this option opens accelerator mode to security
2342 vulnerabilities usually only affecting in interception
2343 mode. Make sure to protect forwarding with suitable
2344 http_access rules when using this.
2347 SSL Bump Mode Options:
2348 In addition to these options ssl-bump requires TLS/SSL options.
2350 generate-host-certificates[=<on|off>]
2351 Dynamically create SSL server certificates for the
2352 destination hosts of bumped CONNECT requests.When
2353 enabled, the cert and key options are used to sign
2354 generated certificates. Otherwise generated
2355 certificate will be selfsigned.
2356 If there is a CA certificate lifetime of the generated
2357 certificate equals lifetime of the CA certificate. If
2358 generated certificate is selfsigned lifetime is three
2360 This option is enabled by default when ssl-bump is used.
2361 See the ssl-bump option above for more information.
2363 dynamic_cert_mem_cache_size=SIZE
2364 Approximate total RAM size spent on cached generated
2365 certificates. If set to zero, caching is disabled. The
2366 default value is 4MB.
2370 tls-cert= Path to file containing an X.509 certificate (PEM format)
2371 to be used in the TLS handshake ServerHello.
2373 If this certificate is constrained by KeyUsage TLS
2374 feature it must allow HTTP server usage, along with
2375 any additional restrictions imposed by your choice
2376 of options= settings.
2378 When OpenSSL is used this file may also contain a
2379 chain of intermediate CA certificates to send in the
2382 When GnuTLS is used this option (and any paired
2383 tls-key= option) may be repeated to load multiple
2384 certificates for different domains.
2386 Also, when generate-host-certificates=on is configured
2387 the first tls-cert= option must be a CA certificate
2388 capable of signing the automatically generated
2391 tls-key= Path to a file containing private key file (PEM format)
2392 for the previous tls-cert= option.
2394 If tls-key= is not specified tls-cert= is assumed to
2395 reference a PEM file containing both the certificate
2398 cipher= Colon separated list of supported ciphers.
2399 NOTE: some ciphers such as EDH ciphers depend on
2400 additional settings. If those settings are
2401 omitted the ciphers may be silently ignored
2402 by the OpenSSL library.
2404 options= Various SSL implementation options. The most important
2407 NO_SSLv3 Disallow the use of SSLv3
2409 NO_TLSv1 Disallow the use of TLSv1.0
2411 NO_TLSv1_1 Disallow the use of TLSv1.1
2413 NO_TLSv1_2 Disallow the use of TLSv1.2
2416 Always create a new key when using
2417 temporary/ephemeral DH key exchanges
2420 Enable ephemeral ECDH key exchange.
2421 The adopted curve should be specified
2422 using the tls-dh option.
2425 Disable use of RFC5077 session tickets.
2426 Some servers may have problems
2427 understanding the TLS extension due
2428 to ambiguous specification in RFC4507.
2430 ALL Enable various bug workarounds
2431 suggested as "harmless" by OpenSSL
2432 Be warned that this reduces SSL/TLS
2433 strength to some attacks.
2435 See the OpenSSL SSL_CTX_set_options documentation for a
2438 clientca= File containing the list of CAs to use when
2439 requesting a client certificate.
2441 tls-cafile= PEM file containing CA certificates to use when verifying
2442 client certificates. If not configured clientca will be
2443 used. May be repeated to load multiple files.
2445 capath= Directory containing additional CA certificates
2446 and CRL lists to use when verifying client certificates.
2447 Requires OpenSSL or LibreSSL.
2449 crlfile= File of additional CRL lists to use when verifying
2450 the client certificate, in addition to CRLs stored in
2451 the capath. Implies VERIFY_CRL flag below.
2454 File containing DH parameters for temporary/ephemeral DH key
2455 exchanges, optionally prefixed by a curve for ephemeral ECDH
2457 See OpenSSL documentation for details on how to create the
2458 DH parameter file. Supported curves for ECDH can be listed
2459 using the "openssl ecparam -list_curves" command.
2460 WARNING: EDH and EECDH ciphers will be silently disabled if
2461 this option is not set.
2463 sslflags= Various flags modifying the use of SSL:
2465 Don't request client certificates
2466 immediately, but wait until acl processing
2467 requires a certificate (not yet implemented).
2469 Request a client certificate during the TLS
2470 handshake, but ignore certificate absence in
2471 the TLS client Hello. If the client does
2472 supply a certificate, it is validated.
2474 Don't allow for session reuse. Each connection
2475 will result in a new SSL session.
2477 Verify CRL lists when accepting client
2480 Verify CRL lists for all certificates in the
2481 client certificate chain.
2483 tls-default-ca[=off]
2484 Whether to use the system Trusted CAs. Default is OFF.
2486 tls-no-npn Do not use the TLS NPN extension to advertise HTTP/1.1.
2488 sslcontext= SSL session ID context identifier.
2492 connection-auth[=on|off]
2493 use connection-auth=off to tell Squid to prevent
2494 forwarding Microsoft connection oriented authentication
2495 (NTLM, Negotiate and Kerberos)
2497 disable-pmtu-discovery=
2498 Control Path-MTU discovery usage:
2499 off lets OS decide on what to do (default).
2500 transparent disable PMTU discovery when transparent
2502 always disable always PMTU discovery.
2504 In many setups of transparently intercepting proxies
2505 Path-MTU discovery can not work on traffic towards the
2506 clients. This is the case when the intercepting device
2507 does not fully track connections and fails to forward
2508 ICMP must fragment messages to the cache server. If you
2509 have such setup and experience that certain clients
2510 sporadically hang or never complete requests set
2511 disable-pmtu-discovery option to 'transparent'.
2513 name= Specifies a internal name for the port. Defaults to
2514 the port specification (port or addr:port)
2516 tcpkeepalive[=idle,interval,timeout]
2517 Enable TCP keepalive probes of idle connections.
2518 In seconds; idle is the initial time before TCP starts
2519 probing the connection, interval how often to probe, and
2520 timeout the time before giving up.
2522 require-proxy-header
2523 Require PROXY protocol version 1 or 2 connections.
2524 The proxy_protocol_access is required to permit
2525 downstream proxies which can be trusted.
2528 Ask TCP stack to maintain a dedicated listening queue
2529 for each worker accepting requests at this port.
2530 Requires TCP stack that supports the SO_REUSEPORT socket
2533 SECURITY WARNING: Enabling worker-specific queues
2534 allows any process running as Squid's effective user to
2535 easily accept requests destined to this port.
2537 If you run Squid on a dual-homed machine with an internal
2538 and an external interface we recommend you to specify the
2539 internal address:port in http_port. This way Squid will only be
2540 visible on the internal address.
2544 # Squid normally listens to port 3128
2545 http_port @DEFAULT_HTTP_PORT@
2550 IFDEF: HAVE_LIBGNUTLS||USE_OPENSSL
2555 Usage: [ip:]port [mode] tls-cert=certificate.pem [options]
2557 The socket address where Squid will listen for client requests made
2558 over TLS or SSL connections. Commonly referred to as HTTPS.
2560 This is most useful for situations where you are running squid in
2561 accelerator mode and you want to do the TLS work at the accelerator
2564 You may specify multiple socket addresses on multiple lines,
2565 each with their own certificate and/or options.
2567 The tls-cert= option is mandatory on HTTPS ports.
2569 See http_port for a list of modes and options.
2570 Not all http_port options are available for https_port.
2571 Among the unavalable options:
2572 - require-proxy-header
2580 Enables Native FTP proxy by specifying the socket address where Squid
2581 listens for FTP client requests. See http_port directive for various
2582 ways to specify the listening address and mode.
2584 Usage: ftp_port address [mode] [options]
2586 WARNING: This is a new, experimental, complex feature that has seen
2587 limited production exposure. Some Squid modules (e.g., caching) do not
2588 currently work with native FTP proxying, and many features have not
2589 even been tested for compatibility. Test well before deploying!
2591 Native FTP proxying differs substantially from proxying HTTP requests
2592 with ftp:// URIs because Squid works as an FTP server and receives
2593 actual FTP commands (rather than HTTP requests with FTP URLs).
2595 Native FTP commands accepted at ftp_port are internally converted or
2596 wrapped into HTTP-like messages. The same happens to Native FTP
2597 responses received from FTP origin servers. Those HTTP-like messages
2598 are shoveled through regular access control and adaptation layers
2599 between the FTP client and the FTP origin server. This allows Squid to
2600 examine, adapt, block, and log FTP exchanges. Squid reuses most HTTP
2601 mechanisms when shoveling wrapped FTP messages. For example,
2602 http_access and adaptation_access directives are used.
2606 intercept Same as http_port intercept. The FTP origin address is
2607 determined based on the intended destination of the
2608 intercepted connection.
2610 tproxy Support Linux TPROXY for spoofing outgoing
2611 connections using the client IP address.
2612 NP: disables authentication and maybe IPv6 on the port.
2614 By default (i.e., without an explicit mode option), Squid extracts the
2615 FTP origin address from the login@origin parameter of the FTP USER
2616 command. Many popular FTP clients support such native FTP proxying.
2620 name=token Specifies an internal name for the port. Defaults to
2621 the port address. Usable with myportname ACL.
2624 Enables tracking of FTP directories by injecting extra
2625 PWD commands and adjusting Request-URI (in wrapping
2626 HTTP requests) to reflect the current FTP server
2627 directory. Tracking is disabled by default.
2629 protocol=FTP Protocol to reconstruct accelerated and intercepted
2630 requests with. Defaults to FTP. No other accepted
2631 values have been tested with. An unsupported value
2632 results in a FATAL error. Accepted values are FTP,
2633 HTTP (or HTTP/1.1), and HTTPS (or HTTPS/1.1).
2635 Other http_port modes and options that are not specific to HTTP and
2636 HTTPS may also work.
2637 Among the options that are not available for ftp_port:
2638 - require-proxy-header
2642 NAME: tcp_outgoing_tos tcp_outgoing_ds tcp_outgoing_dscp
2645 LOC: Ip::Qos::TheConfig.tosToServer
2647 Allows you to select a TOS/Diffserv value for packets outgoing
2648 on the server side, based on an ACL.
2650 tcp_outgoing_tos ds-field [!]aclname ...
2652 Example where normal_service_net uses the TOS value 0x00
2653 and good_service_net uses 0x20
2655 acl normal_service_net src 10.0.0.0/24
2656 acl good_service_net src 10.0.1.0/24
2657 tcp_outgoing_tos 0x00 normal_service_net
2658 tcp_outgoing_tos 0x20 good_service_net
2660 TOS/DSCP values really only have local significance - so you should
2661 know what you're specifying. For more information, see RFC2474,
2662 RFC2475, and RFC3260.
2664 The TOS/DSCP byte must be exactly that - a octet value 0 - 255, or
2665 "default" to use whatever default your host has.
2666 Note that only multiples of 4 are usable as the two rightmost bits have
2667 been redefined for use by ECN (RFC 3168 section 23.1).
2668 The squid parser will enforce this by masking away the ECN bits.
2670 Processing proceeds in the order specified, and stops at first fully
2673 Only fast ACLs are supported.
2676 NAME: clientside_tos
2679 LOC: Ip::Qos::TheConfig.tosToClient
2681 Allows you to select a TOS/DSCP value for packets being transmitted
2682 on the client-side, based on an ACL.
2684 clientside_tos ds-field [!]aclname ...
2686 Example where normal_service_net uses the TOS value 0x00
2687 and good_service_net uses 0x20
2689 acl normal_service_net src 10.0.0.0/24
2690 acl good_service_net src 10.0.1.0/24
2691 clientside_tos 0x00 normal_service_net
2692 clientside_tos 0x20 good_service_net
2694 Note: This feature is incompatible with qos_flows. Any TOS values set here
2695 will be overwritten by TOS values in qos_flows.
2697 The TOS/DSCP byte must be exactly that - a octet value 0 - 255, or
2698 "default" to use whatever default your host has.
2699 Note that only multiples of 4 are usable as the two rightmost bits have
2700 been redefined for use by ECN (RFC 3168 section 23.1).
2701 The squid parser will enforce this by masking away the ECN bits.
2703 This clause only supports fast acl types.
2704 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
2707 NAME: tcp_outgoing_mark
2709 IFDEF: HAVE_LIBCAP&&SO_MARK
2711 LOC: Ip::Qos::TheConfig.nfmarkToServer
2713 Allows you to apply a Netfilter mark value to outgoing packets
2714 on the server side, based on an ACL.
2716 tcp_outgoing_mark mark-value [!]aclname ...
2718 Example where normal_service_net uses the mark value 0x00
2719 and good_service_net uses 0x20
2721 acl normal_service_net src 10.0.0.0/24
2722 acl good_service_net src 10.0.1.0/24
2723 tcp_outgoing_mark 0x00 normal_service_net
2724 tcp_outgoing_mark 0x20 good_service_net
2726 Only fast ACLs are supported.
2729 NAME: mark_client_packet clientside_mark
2731 IFDEF: HAVE_LIBCAP&&SO_MARK
2733 LOC: Ip::Qos::TheConfig.nfmarkToClient
2735 Allows you to apply a Netfilter MARK value to packets being transmitted
2736 on the client-side, based on an ACL.
2738 mark_client_packet mark-value [!]aclname ...
2740 Example where normal_service_net uses the MARK value 0x00
2741 and good_service_net uses 0x20
2743 acl normal_service_net src 10.0.0.0/24
2744 acl good_service_net src 10.0.1.0/24
2745 mark_client_packet 0x00 normal_service_net
2746 mark_client_packet 0x20 good_service_net
2748 Note: This feature is incompatible with qos_flows. Any mark values set here
2749 will be overwritten by mark values in qos_flows.
2751 This clause only supports fast acl types.
2752 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
2755 NAME: mark_client_connection
2757 IFDEF: HAVE_LIBCAP&&SO_MARK
2759 LOC: Ip::Qos::TheConfig.nfConnmarkToClient
2761 Allows you to apply a Netfilter CONNMARK value to a connection
2762 on the client-side, based on an ACL.
2764 mark_client_connection mark-value[/mask] [!]aclname ...
2766 The mark-value and mask are unsigned integers (hex, octal, or decimal).
2767 The mask may be used to preserve marking previously set by other agents
2770 A matching rule replaces the CONNMARK value. If a mask is also
2771 specified, then the masked bits of the original value are zeroed, and
2772 the configured mark-value is ORed with that adjusted value.
2773 For example, applying a mark-value 0xAB/0xF to 0x5F CONNMARK, results
2774 in a 0xFB marking (rather than a 0xAB or 0x5B).
2776 This directive semantics is similar to iptables --set-mark rather than
2777 --set-xmark functionality.
2779 The directive does not interfere with qos_flows (which uses packet MARKs,
2782 Example where squid marks intercepted FTP connections:
2784 acl proto_ftp proto FTP
2785 mark_client_connection 0x200/0xff00 proto_ftp
2787 This clause only supports fast acl types.
2788 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
2795 LOC: Ip::Qos::TheConfig
2797 Allows you to select a TOS/DSCP value to mark outgoing
2798 connections to the client, based on where the reply was sourced.
2799 For platforms using netfilter, allows you to set a netfilter mark
2800 value instead of, or in addition to, a TOS value.
2802 By default this functionality is disabled. To enable it with the default
2803 settings simply use "qos_flows mark" or "qos_flows tos". Default
2804 settings will result in the netfilter mark or TOS value being copied
2805 from the upstream connection to the client. Note that it is the connection
2806 CONNMARK value not the packet MARK value that is copied.
2808 It is not currently possible to copy the mark or TOS value from the
2809 client to the upstream connection request.
2811 TOS values really only have local significance - so you should
2812 know what you're specifying. For more information, see RFC2474,
2813 RFC2475, and RFC3260.
2815 The TOS/DSCP byte must be exactly that - a octet value 0 - 255.
2816 Note that only multiples of 4 are usable as the two rightmost bits have
2817 been redefined for use by ECN (RFC 3168 section 23.1).
2818 The squid parser will enforce this by masking away the ECN bits.
2820 Mark values can be any unsigned 32-bit integer value.
2822 This setting is configured by setting the following values:
2824 tos|mark Whether to set TOS or netfilter mark values
2826 local-hit=0xFF Value to mark local cache hits.
2828 sibling-hit=0xFF Value to mark hits from sibling peers.
2830 parent-hit=0xFF Value to mark hits from parent peers.
2832 miss=0xFF[/mask] Value to mark cache misses. Takes precedence
2833 over the preserve-miss feature (see below), unless
2834 mask is specified, in which case only the bits
2835 specified in the mask are written.
2837 The TOS variant of the following features are only possible on Linux
2838 and require your kernel to be patched with the TOS preserving ZPH
2839 patch, available from http://zph.bratcheda.org
2840 No patch is needed to preserve the netfilter mark, which will work
2841 with all variants of netfilter.
2843 disable-preserve-miss
2844 This option disables the preservation of the TOS or netfilter
2845 mark. By default, the existing TOS or netfilter mark value of
2846 the response coming from the remote server will be retained
2847 and masked with miss-mark.
2848 NOTE: in the case of a netfilter mark, the mark must be set on
2849 the connection (using the CONNMARK target) not on the packet
2853 Allows you to mask certain bits in the TOS or mark value
2854 received from the remote server, before copying the value to
2855 the TOS sent towards clients.
2856 Default for tos: 0xFF (TOS from server is not changed).
2857 Default for mark: 0xFFFFFFFF (mark from server is not changed).
2859 All of these features require the --enable-zph-qos compilation flag
2860 (enabled by default). Netfilter marking also requires the
2861 libnetfilter_conntrack libraries (--with-netfilter-conntrack) and
2862 libcap 2.09+ (--with-libcap).
2866 NAME: tcp_outgoing_address
2869 DEFAULT_DOC: Address selection is performed by the operating system.
2870 LOC: Config.accessList.outgoing_address
2872 Allows you to map requests to different outgoing IP addresses
2873 based on the username or source address of the user making
2876 tcp_outgoing_address ipaddr [[!]aclname] ...
2879 Forwarding clients with dedicated IPs for certain subnets.
2881 acl normal_service_net src 10.0.0.0/24
2882 acl good_service_net src 10.0.2.0/24
2884 tcp_outgoing_address 2001:db8::c001 good_service_net
2885 tcp_outgoing_address 10.1.0.2 good_service_net
2887 tcp_outgoing_address 2001:db8::beef normal_service_net
2888 tcp_outgoing_address 10.1.0.1 normal_service_net
2890 tcp_outgoing_address 2001:db8::1
2891 tcp_outgoing_address 10.1.0.3
2893 Processing proceeds in the order specified, and stops at first fully
2896 Squid will add an implicit IP version test to each line.
2897 Requests going to IPv4 websites will use the outgoing 10.1.0.* addresses.
2898 Requests going to IPv6 websites will use the outgoing 2001:db8:* addresses.
2901 NOTE: The use of this directive using client dependent ACLs is
2902 incompatible with the use of server side persistent connections. To
2903 ensure correct results it is best to set server_persistent_connections
2904 to off when using this directive in such configurations.
2906 NOTE: The use of this directive to set a local IP on outgoing TCP links
2907 is incompatible with using TPROXY to set client IP out outbound TCP links.
2908 When needing to contact peers use the no-tproxy cache_peer option and the
2909 client_dst_passthru directive re-enable normal forwarding such as this.
2911 This clause only supports fast acl types.
2912 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
2915 NAME: host_verify_strict
2918 LOC: Config.onoff.hostStrictVerify
2920 Regardless of this option setting, when dealing with intercepted
2921 traffic, Squid always verifies that the destination IP address matches
2922 the Host header domain or IP (called 'authority form URL').
2924 This enforcement is performed to satisfy a MUST-level requirement in
2925 RFC 2616 section 14.23: "The Host field value MUST represent the naming
2926 authority of the origin server or gateway given by the original URL".
2929 Squid always responds with an HTTP 409 (Conflict) error
2930 page and logs a security warning if there is no match.
2932 Squid verifies that the destination IP address matches
2933 the Host header for forward-proxy and reverse-proxy traffic
2934 as well. For those traffic types, Squid also enables the
2935 following checks, comparing the corresponding Host header
2936 and Request-URI components:
2938 * The host names (domain or IP) must be identical,
2939 but valueless or missing Host header disables all checks.
2940 For the two host names to match, both must be either IP
2943 * Port numbers must be identical, but if a port is missing
2944 the scheme-default port is assumed.
2947 When set to OFF (the default):
2948 Squid allows suspicious requests to continue but logs a
2949 security warning and blocks caching of the response.
2951 * Forward-proxy traffic is not checked at all.
2953 * Reverse-proxy traffic is not checked at all.
2955 * Intercepted traffic which passes verification is handled
2956 according to client_dst_passthru.
2958 * Intercepted requests which fail verification are sent
2959 to the client original destination instead of DIRECT.
2960 This overrides 'client_dst_passthru off'.
2962 For now suspicious intercepted CONNECT requests are always
2963 responded to with an HTTP 409 (Conflict) error page.
2968 As described in CVE-2009-0801 when the Host: header alone is used
2969 to determine the destination of a request it becomes trivial for
2970 malicious scripts on remote websites to bypass browser same-origin
2971 security policy and sandboxing protections.
2973 The cause of this is that such applets are allowed to perform their
2974 own HTTP stack, in which case the same-origin policy of the browser
2975 sandbox only verifies that the applet tries to contact the same IP
2976 as from where it was loaded at the IP level. The Host: header may
2977 be different from the connected IP and approved origin.
2981 NAME: client_dst_passthru
2984 LOC: Config.onoff.client_dst_passthru
2986 With NAT or TPROXY intercepted traffic Squid may pass the request
2987 directly to the original client destination IP or seek a faster
2988 source using the HTTP Host header.
2990 Using Host to locate alternative servers can provide faster
2991 connectivity with a range of failure recovery options.
2992 But can also lead to connectivity trouble when the client and
2993 server are attempting stateful interactions unaware of the proxy.
2995 This option (on by default) prevents alternative DNS entries being
2996 located to send intercepted traffic DIRECT to an origin server.
2997 The clients original destination IP and port will be used instead.
2999 Regardless of this option setting, when dealing with intercepted
3000 traffic Squid will verify the Host: header and any traffic which
3001 fails Host verification will be treated as if this option were ON.
3003 see host_verify_strict for details on the verification process.
3008 -----------------------------------------------------------------------------
3011 NAME: tls_outgoing_options
3012 IFDEF: HAVE_LIBGNUTLS||USE_OPENSSL
3013 TYPE: securePeerOptions
3014 DEFAULT: min-version=1.0
3015 LOC: Security::ProxyOutgoingConfig()
3017 disable Do not support https:// URLs.
3019 cert=/path/to/client/certificate
3020 A client X.509 certificate to use when connecting.
3022 key=/path/to/client/private_key
3023 The private key corresponding to the cert= above.
3025 If key= is not specified cert= is assumed to
3026 reference a PEM file containing both the certificate
3029 cipher=... The list of valid TLS ciphers to use.
3032 The minimum TLS protocol version to permit.
3033 To control SSLv3 use the options= parameter.
3034 Supported Values: 1.0 (default), 1.1, 1.2, 1.3
3036 options=... Specify various TLS/SSL implementation options.
3038 OpenSSL options most important are:
3040 NO_SSLv3 Disallow the use of SSLv3
3043 Always create a new key when using
3044 temporary/ephemeral DH key exchanges
3047 Disable use of RFC5077 session tickets.
3048 Some servers may have problems
3049 understanding the TLS extension due
3050 to ambiguous specification in RFC4507.
3052 ALL Enable various bug workarounds
3053 suggested as "harmless" by OpenSSL
3054 Be warned that this reduces SSL/TLS
3055 strength to some attacks.
3057 See the OpenSSL SSL_CTX_set_options documentation
3058 for a more complete list.
3060 GnuTLS options most important are:
3063 Disable use of RFC5077 session tickets.
3064 Some servers may have problems
3065 understanding the TLS extension due
3066 to ambiguous specification in RFC4507.
3068 See the GnuTLS Priority Strings documentation
3069 for a more complete list.
3070 http://www.gnutls.org/manual/gnutls.html#Priority-Strings
3073 cafile= PEM file containing CA certificates to use when verifying
3074 the peer certificate. May be repeated to load multiple files.
3076 capath= A directory containing additional CA certificates to
3077 use when verifying the peer certificate.
3078 Requires OpenSSL or LibreSSL.
3080 crlfile=... A certificate revocation list file to use when
3081 verifying the peer certificate.
3083 flags=... Specify various flags modifying the TLS implementation:
3086 Accept certificates even if they fail to
3089 Don't verify the peer certificate
3090 matches the server name
3093 Whether to use the system Trusted CAs. Default is ON.
3095 domain= The peer name as advertised in its certificate.
3096 Used for verifying the correctness of the received peer
3097 certificate. If not specified the peer hostname will be
3103 -----------------------------------------------------------------------------
3106 NAME: ssl_unclean_shutdown
3110 LOC: Config.SSL.unclean_shutdown
3112 Some browsers (especially MSIE) bugs out on SSL shutdown
3119 LOC: Config.SSL.ssl_engine
3122 The OpenSSL engine to use. You will need to set this if you
3123 would like to use hardware SSL acceleration for example.
3125 Not supported in builds with OpenSSL 3.0 or newer.
3128 NAME: sslproxy_session_ttl
3131 LOC: Config.SSL.session_ttl
3134 Sets the timeout value for SSL sessions
3137 NAME: sslproxy_session_cache_size
3140 LOC: Config.SSL.sessionCacheSize
3143 Sets the cache size to use for ssl session
3146 NAME: sslproxy_foreign_intermediate_certs
3149 LOC: Config.ssl_client.foreignIntermediateCertsPath
3152 Many origin servers fail to send their full server certificate
3153 chain for verification, assuming the client already has or can
3154 easily locate any missing intermediate certificates.
3156 Squid uses the certificates from the specified file to fill in
3157 these missing chains when trying to validate origin server
3160 The file is expected to contain zero or more PEM-encoded
3161 intermediate certificates. These certificates are not treated
3162 as trusted root certificates, and any self-signed certificate in
3163 this file will be ignored.
3166 NAME: sslproxy_cert_sign_hash
3169 LOC: Config.SSL.certSignHash
3172 Sets the hashing algorithm to use when signing generated certificates.
3173 Valid algorithm names depend on the OpenSSL library used. The following
3174 names are usually available: sha1, sha256, sha512, and md5. Please see
3175 your OpenSSL library manual for the available hashes. By default, Squids
3176 that support this option use sha256 hashes.
3178 Squid does not forcefully purge cached certificates that were generated
3179 with an algorithm other than the currently configured one. They remain
3180 in the cache, subject to the regular cache eviction policy, and become
3181 useful if the algorithm changes again.
3186 TYPE: sslproxy_ssl_bump
3187 LOC: Config.accessList.ssl_bump
3188 DEFAULT_DOC: Become a TCP tunnel without decrypting proxied traffic.
3191 This option is consulted when a CONNECT request is received on
3192 an http_port (or a new connection is intercepted at an
3193 https_port), provided that port was configured with an ssl-bump
3194 flag. The subsequent data on the connection is either treated as
3195 HTTPS and decrypted OR tunneled at TCP level without decryption,
3196 depending on the first matching bumping "action".
3198 ssl_bump <action> [!]acl ...
3200 The following bumping actions are currently supported:
3203 Become a TCP tunnel without decrypting proxied traffic.
3204 This is the default action.
3207 When used on step SslBump1, establishes a secure connection
3208 with the client first, then connect to the server.
3209 When used on step SslBump2 or SslBump3, establishes a secure
3210 connection with the server and, using a mimicked server
3211 certificate, with the client.
3214 Receive client (step SslBump1) or server (step SslBump2)
3215 certificate while preserving the possibility of splicing the
3216 connection. Peeking at the server certificate (during step 2)
3217 usually precludes bumping of the connection at step 3.
3220 Receive client (step SslBump1) or server (step SslBump2)
3221 certificate while preserving the possibility of bumping the
3222 connection. Staring at the server certificate (during step 2)
3223 usually precludes splicing of the connection at step 3.
3226 Close client and server connections.
3228 Backward compatibility actions available at step SslBump1:
3231 Bump the connection. Establish a secure connection with the
3232 client first, then connect to the server. This old mode does
3233 not allow Squid to mimic server SSL certificate and does not
3234 work with intercepted SSL connections.
3237 Bump the connection. Establish a secure connection with the
3238 server first, then establish a secure connection with the
3239 client, using a mimicked server certificate. Works with both
3240 CONNECT requests and intercepted SSL connections, but does
3241 not allow to make decisions based on SSL handshake info.
3244 Decide whether to bump or splice the connection based on
3245 client-to-squid and server-to-squid SSL hello messages.
3249 Same as the "splice" action.
3251 All ssl_bump rules are evaluated at each of the supported bumping
3252 steps. Rules with actions that are impossible at the current step are
3253 ignored. The first matching ssl_bump action wins and is applied at the
3254 end of the current step. If no rules match, the splice action is used.
3255 See the at_step ACL for a list of the supported SslBump steps.
3257 This clause supports both fast and slow acl types.
3258 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
3260 See also: http_port ssl-bump, https_port ssl-bump, and acl at_step.
3263 # Example: Bump all TLS connections except those originating from
3264 # localhost or those going to example.com.
3266 acl broken_sites ssl::server_name .example.com
3267 ssl_bump splice localhost
3268 ssl_bump splice broken_sites
3272 NAME: sslproxy_cert_error
3275 DEFAULT_DOC: Server certificate errors terminate the transaction.
3276 LOC: Config.ssl_client.cert_error
3279 Use this ACL to bypass server certificate validation errors.
3281 For example, the following lines will bypass all validation errors
3282 when talking to servers for example.com. All other
3283 validation errors will result in ERR_SECURE_CONNECT_FAIL error.
3285 acl BrokenButTrustedServers dstdomain example.com
3286 sslproxy_cert_error allow BrokenButTrustedServers
3287 sslproxy_cert_error deny all
3289 This clause only supports fast acl types.
3290 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
3291 Using slow acl types may result in server crashes
3293 Without this option, all server certificate validation errors
3294 terminate the transaction to protect Squid and the client.
3296 SQUID_X509_V_ERR_INFINITE_VALIDATION error cannot be bypassed
3297 but should not happen unless your OpenSSL library is buggy.
3300 Bypassing validation errors is dangerous because an
3301 error usually implies that the server cannot be trusted
3302 and the connection may be insecure.
3304 See also: sslproxy_flags and DONT_VERIFY_PEER.
3307 NAME: sslproxy_cert_sign
3310 POSTSCRIPTUM: signUntrusted ssl::certUntrusted
3311 POSTSCRIPTUM: signSelf ssl::certSelfSigned
3312 POSTSCRIPTUM: signTrusted all
3313 TYPE: sslproxy_cert_sign
3314 LOC: Config.ssl_client.cert_sign
3317 sslproxy_cert_sign <signing algorithm> acl ...
3319 The following certificate signing algorithms are supported:
3322 Sign using the configured CA certificate which is usually
3323 placed in and trusted by end-user browsers. This is the
3324 default for trusted origin server certificates.
3327 Sign to guarantee an X509_V_ERR_CERT_UNTRUSTED browser error.
3328 This is the default for untrusted origin server certificates
3329 that are not self-signed (see ssl::certUntrusted).
3332 Sign using a self-signed certificate with the right CN to
3333 generate a X509_V_ERR_DEPTH_ZERO_SELF_SIGNED_CERT error in the
3334 browser. This is the default for self-signed origin server
3335 certificates (see ssl::certSelfSigned).
3337 This clause only supports fast acl types.
3339 When sslproxy_cert_sign acl(s) match, Squid uses the corresponding
3340 signing algorithm to generate the certificate and ignores all
3341 subsequent sslproxy_cert_sign options (the first match wins). If no
3342 acl(s) match, the default signing algorithm is determined by errors
3343 detected when obtaining and validating the origin server certificate.
3345 WARNING: SQUID_X509_V_ERR_DOMAIN_MISMATCH and ssl:certDomainMismatch can
3346 be used with sslproxy_cert_adapt, but if and only if Squid is bumping a
3347 CONNECT request that carries a domain name. In all other cases (CONNECT
3348 to an IP address or an intercepted SSL connection), Squid cannot detect
3349 the domain mismatch at certificate generation time when
3350 bump-server-first is used.
3353 NAME: sslproxy_cert_adapt
3356 TYPE: sslproxy_cert_adapt
3357 LOC: Config.ssl_client.cert_adapt
3360 sslproxy_cert_adapt <adaptation algorithm> acl ...
3362 The following certificate adaptation algorithms are supported:
3365 Sets the "Not After" property to the "Not After" property of
3366 the CA certificate used to sign generated certificates.
3369 Sets the "Not Before" property to the "Not Before" property of
3370 the CA certificate used to sign generated certificates.
3372 setCommonName or setCommonName{CN}
3373 Sets Subject.CN property to the host name specified as a
3374 CN parameter or, if no explicit CN parameter was specified,
3375 extracted from the CONNECT request. It is a misconfiguration
3376 to use setCommonName without an explicit parameter for
3377 intercepted or tproxied SSL connections.
3379 This clause only supports fast acl types.
3381 Squid first groups sslproxy_cert_adapt options by adaptation algorithm.
3382 Within a group, when sslproxy_cert_adapt acl(s) match, Squid uses the
3383 corresponding adaptation algorithm to generate the certificate and
3384 ignores all subsequent sslproxy_cert_adapt options in that algorithm's
3385 group (i.e., the first match wins within each algorithm group). If no
3386 acl(s) match, the default mimicking action takes place.
3388 WARNING: SQUID_X509_V_ERR_DOMAIN_MISMATCH and ssl:certDomainMismatch can
3389 be used with sslproxy_cert_adapt, but if and only if Squid is bumping a
3390 CONNECT request that carries a domain name. In all other cases (CONNECT
3391 to an IP address or an intercepted SSL connection), Squid cannot detect
3392 the domain mismatch at certificate generation time when
3393 bump-server-first is used.
3396 NAME: sslpassword_program
3399 LOC: Config.Program.ssl_password
3402 Specify a program used for entering SSL key passphrases
3403 when using encrypted SSL certificate keys. If not specified
3404 keys must either be unencrypted, or Squid started with the -N
3405 option to allow it to query interactively for the passphrase.
3407 The key file name is given as argument to the program allowing
3408 selection of the right password if you have multiple encrypted
3413 OPTIONS RELATING TO EXTERNAL SSL_CRTD
3414 -----------------------------------------------------------------------------
3417 NAME: sslcrtd_program
3420 DEFAULT: @DEFAULT_SSL_CRTD@ -s @DEFAULT_SSL_DB_DIR@ -M 4MB
3421 LOC: Ssl::TheConfig.ssl_crtd
3423 Specify the location and options of the executable for certificate
3426 @DEFAULT_SSL_CRTD@ program can use a disk cache to improve response
3427 times on repeated requests. To enable caching, specify -s and -M
3428 parameters. If those parameters are not given, the program generates
3429 a new certificate on every request.
3431 For more information use:
3432 @DEFAULT_SSL_CRTD@ -h
3435 NAME: sslcrtd_children
3436 TYPE: HelperChildConfig
3438 DEFAULT: 32 startup=5 idle=1
3439 LOC: Ssl::TheConfig.ssl_crtdChildren
3441 Specifies the maximum number of certificate generation processes that
3442 Squid may spawn (numberofchildren) and several related options. Using
3443 too few of these helper processes (a.k.a. "helpers") creates request
3444 queues. Using too many helpers wastes your system resources. Squid
3445 does not support spawning more than 32 helpers.
3447 Usage: numberofchildren [option]...
3449 The startup= and idle= options allow some measure of skew in your
3454 Sets the minimum number of processes to spawn when Squid
3455 starts or reconfigures. When set to zero the first request will
3456 cause spawning of the first child process to handle it.
3458 Starting too few children temporary slows Squid under load while it
3459 tries to spawn enough additional processes to cope with traffic.
3463 Sets a minimum of how many processes Squid is to try and keep available
3464 at all times. When traffic begins to rise above what the existing
3465 processes can handle this many more will be spawned up to the maximum
3466 configured. A minimum setting of 1 is required.
3470 Sets the maximum number of queued requests. A request is queued when
3471 no existing child is idle and no new child can be started due to
3472 numberofchildren limit. If the queued requests exceed queue size for
3473 more than 3 minutes squid aborts its operation. The default value is
3474 set to 2*numberofchildren.
3476 You must have at least one ssl_crtd process.
3479 NAME: sslcrtvalidator_program
3483 LOC: Ssl::TheConfig.ssl_crt_validator
3485 Specify the location and options of the executable for ssl_crt_validator
3488 Usage: sslcrtvalidator_program [ttl=...] [cache=n] path ...
3493 Limits how much memory Squid can use for caching validator
3494 responses. The default is 67108864 (i.e. 64 MB).
3495 Reconfiguration purges any excess entries. To disable caching,
3496 use cache=0. Currently, cache entry sizes are seriously
3497 underestimated. Even with that bug, a typical estimate for a
3498 single cache entry size would be at least a few kilobytes (the
3499 size of the PEM certificates sent to the validator).
3501 ttl=<seconds|"infinity">
3502 Approximately how long Squid may reuse the validator results
3503 for. The default is 3600 (i.e. 1 hour). Using ttl=infinity
3504 disables TTL checks. Reconfiguration does not affect TTLs of
3505 the already cached entries. To disable caching, use zero cache
3506 size, not zero TTL -- zero TTL allows reuse for the remainder
3507 of the second when the result was cached.
3510 NAME: sslcrtvalidator_children
3511 TYPE: HelperChildConfig
3513 DEFAULT: 32 startup=5 idle=1 concurrency=1
3514 LOC: Ssl::TheConfig.ssl_crt_validator_Children
3516 Specifies the maximum number of certificate validation processes that
3517 Squid may spawn (numberofchildren) and several related options. Using
3518 too few of these helper processes (a.k.a. "helpers") creates request
3519 queues. Using too many helpers wastes your system resources. Squid
3520 does not support spawning more than 32 helpers.
3522 Usage: numberofchildren [option]...
3524 The startup= and idle= options allow some measure of skew in your
3529 Sets the minimum number of processes to spawn when Squid
3530 starts or reconfigures. When set to zero the first request will
3531 cause spawning of the first child process to handle it.
3533 Starting too few children temporary slows Squid under load while it
3534 tries to spawn enough additional processes to cope with traffic.
3538 Sets a minimum of how many processes Squid is to try and keep available
3539 at all times. When traffic begins to rise above what the existing
3540 processes can handle this many more will be spawned up to the maximum
3541 configured. A minimum setting of 1 is required.
3545 The number of requests each certificate validator helper can handle in
3546 parallel. A value of 0 indicates the certificate validator does not
3547 support concurrency. Defaults to 1.
3549 When this directive is set to a value >= 1 then the protocol
3550 used to communicate with the helper is modified to include
3551 a request ID in front of the request/response. The request
3552 ID from the request must be echoed back with the response
3557 Sets the maximum number of queued requests. A request is queued when
3558 no existing child can accept it due to concurrency limit and no new
3559 child can be started due to numberofchildren limit. If the queued
3560 requests exceed queue size for more than 3 minutes squid aborts its
3561 operation. The default value is set to 2*numberofchildren.
3563 You must have at least one ssl_crt_validator process.
3567 OPTIONS WHICH AFFECT THE NEIGHBOR SELECTION ALGORITHM
3568 -----------------------------------------------------------------------------
3576 To specify other caches in a hierarchy, use the format:
3578 cache_peer hostname type http-port icp-port [options]
3583 # hostname type port port options
3584 # -------------------- -------- ----- ----- -----------
3585 cache_peer parent.foo.net parent 3128 3130 default
3586 cache_peer sib1.foo.net sibling 3128 3130 proxy-only
3587 cache_peer sib2.foo.net sibling 3128 3130 proxy-only
3588 cache_peer example.com parent 80 0 default
3589 cache_peer cdn.example.com sibling 3128 0
3591 type: either 'parent', 'sibling', or 'multicast'.
3593 proxy-port: The port number where the peer accept HTTP requests.
3594 For other Squid proxies this is usually 3128
3595 For web servers this is usually 80
3597 icp-port: Used for querying neighbor caches about objects.
3598 Set to 0 if the peer does not support ICP or HTCP.
3599 See ICP and HTCP options below for additional details.
3602 ==== ICP OPTIONS ====
3604 You MUST also set icp_port and icp_access explicitly when using these options.
3605 The defaults will prevent peer traffic using ICP.
3608 no-query Disable ICP queries to this neighbor.
3611 Indicates the named peer is a member of a multicast group.
3612 ICP queries will not be sent directly to the peer, but ICP
3613 replies will be accepted from it.
3615 closest-only Indicates that, for ICP_OP_MISS replies, we'll only forward
3616 CLOSEST_PARENT_MISSes and never FIRST_PARENT_MISSes.
3619 To only send ICP queries to this neighbor infrequently.
3620 This is used to keep the neighbor round trip time updated
3621 and is usually used in conjunction with weighted-round-robin.
3624 ==== HTCP OPTIONS ====
3626 You MUST also set htcp_port and htcp_access explicitly when using these options.
3627 The defaults will prevent peer traffic using HTCP.
3630 htcp Send HTCP, instead of ICP, queries to the neighbor.
3631 You probably also want to set the "icp-port" to 4827
3632 instead of 3130. This directive accepts a comma separated
3633 list of options described below.
3635 htcp=oldsquid Send HTCP to old Squid versions (2.5 or earlier).
3637 htcp=no-clr Send HTCP to the neighbor but without
3638 sending any CLR requests. This cannot be used with
3641 htcp=only-clr Send HTCP to the neighbor but ONLY CLR requests.
3642 This cannot be used with no-clr.
3645 Send HTCP to the neighbor including CLRs but only when
3646 they do not result from PURGE requests.
3649 Forward any HTCP CLR requests this proxy receives to the peer.
3652 ==== PEER SELECTION METHODS ====
3654 The default peer selection method is ICP, with the first responding peer
3655 being used as source. These options can be used for better load balancing.
3658 default This is a parent cache which can be used as a "last-resort"
3659 if a peer cannot be located by any of the peer-selection methods.
3660 If specified more than once, only the first is used.
3662 round-robin Load-Balance parents which should be used in a round-robin
3663 fashion in the absence of any ICP queries.
3664 weight=N can be used to add bias.
3666 weighted-round-robin
3667 Load-Balance parents which should be used in a round-robin
3668 fashion with the frequency of each parent being based on the
3669 round trip time. Closer parents are used more often.
3670 Usually used for background-ping parents.
3671 weight=N can be used to add bias.
3673 carp Load-Balance parents which should be used as a CARP array.
3674 The requests will be distributed among the parents based on the
3675 CARP load balancing hash function based on their weight.
3677 userhash Load-balance parents based on the client proxy_auth username.
3679 sourcehash Load-balance parents based on the client source IP.
3682 To be used only for cache peers of type "multicast".
3683 ALL members of this multicast group have "sibling"
3684 relationship with it, not "parent". This is to a multicast
3685 group when the requested object would be fetched only from
3686 a "parent" cache, anyway. It's useful, e.g., when
3687 configuring a pool of redundant Squid proxies, being
3688 members of the same multicast group.
3691 ==== PEER SELECTION OPTIONS ====
3693 weight=N use to affect the selection of a peer during any weighted
3694 peer-selection mechanisms.
3695 The weight must be an integer; default is 1,
3696 larger weights are favored more.
3697 This option does not affect parent selection if a peering
3698 protocol is not in use.
3700 basetime=N Specify a base amount to be subtracted from round trip
3702 It is subtracted before division by weight in calculating
3703 which parent to fectch from. If the rtt is less than the
3704 base time the rtt is set to a minimal value.
3706 ttl=N Specify a TTL to use when sending multicast ICP queries
3708 Only useful when sending to a multicast group.
3709 Because we don't accept ICP replies from random
3710 hosts, you must configure other group members as
3711 peers with the 'multicast-responder' option.
3713 no-delay To prevent access to this neighbor from influencing the
3716 digest-url=URL Tell Squid to fetch the cache digest (if digests are
3717 enabled) for this host from the specified URL rather
3718 than the Squid default location.
3721 ==== CARP OPTIONS ====
3723 carp-key=key-specification
3724 use a different key than the full URL to hash against the peer.
3725 the key-specification is a comma-separated list of the keywords
3726 scheme, host, port, path, params
3727 Order is not important.
3729 ==== ACCELERATOR / REVERSE-PROXY OPTIONS ====
3731 originserver Causes this parent to be contacted as an origin server.
3732 Meant to be used in accelerator setups when the peer
3736 Set the Host header of requests forwarded to this peer.
3737 Useful in accelerator setups where the server (peer)
3738 expects a certain domain name but clients may request
3739 others. ie example.com or www.example.com
3741 no-digest Disable request of cache digests.
3744 Disables requesting ICMP RTT database (NetDB).
3747 ==== AUTHENTICATION OPTIONS ====
3750 If this is a personal/workgroup proxy and your parent
3751 requires proxy authentication.
3753 Note: The string can include URL escapes (i.e. %20 for
3754 spaces). This also means % must be written as %%.
3757 Send login details received from client to this peer.
3758 Both Proxy- and WWW-Authorization headers are passed
3759 without alteration to the peer.
3760 Authentication is not required by Squid for this to work.
3762 Note: This will pass any form of authentication but
3763 only Basic auth will work through a proxy unless the
3764 connection-auth options are also used.
3766 login=PASS Send login details received from client to this peer.
3767 Authentication is not required by this option.
3769 If there are no client-provided authentication headers
3770 to pass on, but username and password are available
3771 from an external ACL user= and password= result tags
3772 they may be sent instead.
3774 Note: To combine this with proxy_auth both proxies must
3775 share the same user database as HTTP only allows for
3776 a single login (one for proxy, one for origin server).
3777 Also be warned this will expose your users proxy
3778 password to the peer. USE WITH CAUTION
3781 Send the username to the upstream cache, but with a
3782 fixed password. This is meant to be used when the peer
3783 is in another administrative domain, but it is still
3784 needed to identify each user.
3785 The star can optionally be followed by some extra
3786 information which is added to the username. This can
3787 be used to identify this proxy to the peer, similar to
3788 the login=username:password option above.
3791 If this is a personal/workgroup proxy and your parent
3792 requires a secure proxy authentication.
3793 The first principal from the default keytab or defined by
3794 the environment variable KRB5_KTNAME will be used.
3796 WARNING: The connection may transmit requests from multiple
3797 clients. Negotiate often assumes end-to-end authentication
3798 and a single-client. Which is not strictly true here.
3800 login=NEGOTIATE:principal_name
3801 If this is a personal/workgroup proxy and your parent
3802 requires a secure proxy authentication.
3803 The principal principal_name from the default keytab or
3804 defined by the environment variable KRB5_KTNAME will be
3807 WARNING: The connection may transmit requests from multiple
3808 clients. Negotiate often assumes end-to-end authentication
3809 and a single-client. Which is not strictly true here.
3811 connection-auth=on|off
3812 Tell Squid that this peer does or not support Microsoft
3813 connection oriented authentication, and any such
3814 challenges received from there should be ignored.
3815 Default is auto to automatically determine the status
3819 Do not use a keytab to authenticate to a peer when
3820 login=NEGOTIATE is specified. Let the GSSAPI
3821 implementation determine which already existing
3822 credentials cache to use instead.
3825 ==== SSL / HTTPS / TLS OPTIONS ====
3827 tls Encrypt connections to this peer with TLS.
3829 sslcert=/path/to/ssl/certificate
3830 A client X.509 certificate to use when connecting to
3833 sslkey=/path/to/ssl/key
3834 The private key corresponding to sslcert above.
3836 If sslkey= is not specified sslcert= is assumed to
3837 reference a PEM file containing both the certificate
3840 sslcipher=... The list of valid SSL ciphers to use when connecting
3844 The minimum TLS protocol version to permit. To control
3845 SSLv3 use the tls-options= parameter.
3846 Supported Values: 1.0 (default), 1.1, 1.2
3848 tls-options=... Specify various TLS implementation options.
3850 OpenSSL options most important are:
3852 NO_SSLv3 Disallow the use of SSLv3
3855 Always create a new key when using
3856 temporary/ephemeral DH key exchanges
3859 Disable use of RFC5077 session tickets.
3860 Some servers may have problems
3861 understanding the TLS extension due
3862 to ambiguous specification in RFC4507.
3864 ALL Enable various bug workarounds
3865 suggested as "harmless" by OpenSSL
3866 Be warned that this reduces SSL/TLS
3867 strength to some attacks.
3869 See the OpenSSL SSL_CTX_set_options documentation for a
3872 GnuTLS options most important are:
3875 Disable use of RFC5077 session tickets.
3876 Some servers may have problems
3877 understanding the TLS extension due
3878 to ambiguous specification in RFC4507.
3880 See the GnuTLS Priority Strings documentation
3881 for a more complete list.
3882 http://www.gnutls.org/manual/gnutls.html#Priority-Strings
3884 tls-cafile= PEM file containing CA certificates to use when verifying
3885 the peer certificate. May be repeated to load multiple files.
3887 sslcapath=... A directory containing additional CA certificates to
3888 use when verifying the peer certificate.
3889 Requires OpenSSL or LibreSSL.
3891 sslcrlfile=... A certificate revocation list file to use when
3892 verifying the peer certificate.
3894 sslflags=... Specify various flags modifying the SSL implementation:
3897 Accept certificates even if they fail to
3901 Don't verify the peer certificate
3902 matches the server name
3904 ssldomain= The peer name as advertised in it's certificate.
3905 Used for verifying the correctness of the received peer
3906 certificate. If not specified the peer hostname will be
3909 front-end-https[=off|on|auto]
3910 Enable the "Front-End-Https: On" header needed when
3911 using Squid as a SSL frontend in front of Microsoft OWA.
3912 See MS KB document Q307347 for details on this header.
3913 If set to auto the header will only be added if the
3914 request is forwarded as a https:// URL.
3916 tls-default-ca[=off]
3917 Whether to use the system Trusted CAs. Default is ON.
3919 tls-no-npn Do not use the TLS NPN extension to advertise HTTP/1.1.
3921 ==== GENERAL OPTIONS ====
3924 A peer-specific connect timeout.
3925 Also see the peer_connect_timeout directive.
3927 connect-fail-limit=N
3928 How many times connecting to a peer must fail before
3929 it is marked as down. Standby connection failures
3930 count towards this limit. Default is 10.
3932 allow-miss Disable Squid's use of only-if-cached when forwarding
3933 requests to siblings. This is primarily useful when
3934 icp_hit_stale is used by the sibling. Excessive use
3935 of this option may result in forwarding loops. One way
3936 to prevent peering loops when using this option, is to
3937 deny cache peer usage on requests from a peer:
3939 cache_peer_access peerName deny fromPeer
3941 max-conn=N Limit the number of concurrent connections the Squid
3942 may open to this peer, including already opened idle
3943 and standby connections. There is no peer-specific
3944 connection limit by default.
3946 A peer exceeding the limit is not used for new
3947 requests unless a standby connection is available.
3949 max-conn currently works poorly with idle persistent
3950 connections: When a peer reaches its max-conn limit,
3951 and there are idle persistent connections to the peer,
3952 the peer may not be selected because the limiting code
3953 does not know whether Squid can reuse those idle
3956 standby=N Maintain a pool of N "hot standby" connections to an
3957 UP peer, available for requests when no idle
3958 persistent connection is available (or safe) to use.
3959 By default and with zero N, no such pool is maintained.
3960 N must not exceed the max-conn limit (if any).
3962 At start or after reconfiguration, Squid opens new TCP
3963 standby connections until there are N connections
3964 available and then replenishes the standby pool as
3965 opened connections are used up for requests. A used
3966 connection never goes back to the standby pool, but
3967 may go to the regular idle persistent connection pool
3968 shared by all peers and origin servers.
3970 Squid never opens multiple new standby connections
3971 concurrently. This one-at-a-time approach minimizes
3972 flooding-like effect on peers. Furthermore, just a few
3973 standby connections should be sufficient in most cases
3974 to supply most new requests with a ready-to-use
3977 Standby connections obey server_idle_pconn_timeout.
3978 For the feature to work as intended, the peer must be
3979 configured to accept and keep them open longer than
3980 the idle timeout at the connecting Squid, to minimize
3981 race conditions typical to idle used persistent
3982 connections. Default request_timeout and
3983 server_idle_pconn_timeout values ensure such a
3986 name=xxx Unique name for the peer.
3987 Required if you have multiple cache_peers with the same hostname.
3988 Defaults to cache_peer hostname when not explicitly specified.
3990 Other directives (e.g., cache_peer_access), cache manager reports,
3991 and cache.log messages use this name to refer to this cache_peer.
3993 The cache_peer name value affects hashing-based peer selection
3994 methods (e.g., carp and sourcehash).
3996 Can be used by outgoing access controls through the
3999 The name value preserves configured spelling, but name uniqueness
4000 checks and name-based search are case-insensitive.
4002 no-tproxy Do not use the client-spoof TPROXY support when forwarding
4003 requests to this peer. Use normal address selection instead.
4004 This overrides the spoof_client_ip ACL.
4006 proxy-only objects fetched from the peer will not be stored locally.
4010 NAME: cache_peer_access
4013 DEFAULT_DOC: No peer usage restrictions.
4016 Restricts usage of cache_peer proxies.
4019 cache_peer_access peer-name allow|deny [!]aclname ...
4021 For the required peer-name parameter, use either the value of the
4022 cache_peer name=value parameter or, if name=value is missing, the
4023 cache_peer hostname parameter.
4025 This directive narrows down the selection of peering candidates, but
4026 does not determine the order in which the selected candidates are
4027 contacted. That order is determined by the peer selection algorithms
4028 (see PEER SELECTION sections in the cache_peer documentation).
4030 If a deny rule matches, the corresponding peer will not be contacted
4031 for the current transaction -- Squid will not send ICP queries and
4032 will not forward HTTP requests to that peer. An allow match leaves
4033 the corresponding peer in the selection. The first match for a given
4034 peer wins for that peer.
4036 The relative order of cache_peer_access directives for the same peer
4037 matters. The relative order of any two cache_peer_access directives
4038 for different peers does not matter. To ease interpretation, it is a
4039 good idea to group cache_peer_access directives for the same peer
4042 A single cache_peer_access directive may be evaluated multiple times
4043 for a given transaction because individual peer selection algorithms
4044 may check it independently from each other. These redundant checks
4045 may be optimized away in future Squid versions.
4047 This clause only supports fast acl types.
4048 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
4052 NAME: neighbor_type_domain
4053 TYPE: hostdomaintype
4055 DEFAULT_DOC: The peer type from cache_peer directive is used for all requests to that peer.
4058 Modify the cache_peer neighbor type when passing requests
4059 about specific domains to the peer.
4062 neighbor_type_domain peer-name parent|sibling domain...
4064 For the required peer-name parameter, use either the value of the
4065 cache_peer name=value parameter or, if name=value is missing, the
4066 cache_peer hostname parameter.
4069 cache_peer foo.example.com parent 3128 3130
4070 neighbor_type_domain foo.example.com sibling .au .de
4072 The above configuration treats all requests to foo.example.com as a
4073 parent proxy unless the request is for a .au or .de ccTLD domain name.
4076 NAME: dead_peer_timeout
4080 LOC: Config.Timeout.deadPeer
4082 This controls how long Squid waits to declare a peer cache
4083 as "dead." If there are no ICP replies received in this
4084 amount of time, Squid will declare the peer dead and not
4085 expect to receive any further ICP replies. However, it
4086 continues to send ICP queries, and will mark the peer as
4087 alive upon receipt of the first subsequent ICP reply.
4089 This timeout also affects when Squid expects to receive ICP
4090 replies from peers. If more than 'dead_peer' seconds have
4091 passed since the last ICP reply was received, Squid will not
4092 expect to receive an ICP reply on the next query. Thus, if
4093 your time between requests is greater than this timeout, you
4094 will see a lot of requests sent DIRECT to origin servers
4095 instead of to your parents.
4098 NAME: forward_max_tries
4101 LOC: Config.forward_max_tries
4103 Limits the number of attempts to forward the request.
4105 For the purpose of this limit, Squid counts all high-level request
4106 forwarding attempts, including any same-destination retries after
4107 certain persistent connection failures and any attempts to use a
4108 different peer. However, these low-level attempts are not counted:
4109 * connection reopening attempts (enabled using connect_retries)
4110 * unfinished Happy Eyeballs connection attempts (prevented by setting
4111 happy_eyeballs_connect_limit to 0)
4113 See also: forward_timeout, connect_retries, and %request_attempts.
4117 MEMORY CACHE OPTIONS
4118 -----------------------------------------------------------------------------
4125 LOC: Config.memMaxSize
4127 NOTE: THIS PARAMETER DOES NOT SPECIFY THE MAXIMUM PROCESS SIZE.
4128 IT ONLY PLACES A LIMIT ON HOW MUCH ADDITIONAL MEMORY SQUID WILL
4129 USE AS A MEMORY CACHE OF OBJECTS. SQUID USES MEMORY FOR OTHER
4130 THINGS AS WELL. SEE THE SQUID FAQ SECTION 8 FOR DETAILS.
4132 'cache_mem' specifies the ideal amount of memory to be used
4134 * In-Transit objects
4136 * Negative-Cached objects
4138 Data for these objects are stored in 4 KB blocks. This
4139 parameter specifies the ideal upper limit on the total size of
4140 4 KB blocks allocated. In-Transit objects take the highest
4143 In-transit objects have priority over the others. When
4144 additional space is needed for incoming data, negative-cached
4145 and hot objects will be released. In other words, the
4146 negative-cached and hot objects will fill up any unused space
4147 not needed for in-transit objects.
4149 If circumstances require, this limit will be exceeded.
4150 Specifically, if your incoming request rate requires more than
4151 'cache_mem' of memory to hold in-transit objects, Squid will
4152 exceed this limit to satisfy the new requests. When the load
4153 decreases, blocks will be freed until the high-water mark is
4154 reached. Thereafter, blocks will be used to store hot
4157 If shared memory caching is enabled, Squid does not use the shared
4158 cache space for in-transit objects, but they still consume as much
4159 local memory as they need. For more details about the shared memory
4160 cache, see memory_cache_shared.
4163 NAME: maximum_object_size_in_memory
4167 LOC: Config.Store.maxInMemObjSize
4169 Objects greater than this size will not be attempted to kept in
4170 the memory cache. This should be set high enough to keep objects
4171 accessed frequently in memory to improve performance whilst low
4172 enough to keep larger objects from hoarding cache_mem.
4175 NAME: memory_cache_shared
4178 LOC: Config.memShared
4180 DEFAULT_DOC: "on" where supported if doing memory caching with multiple SMP workers.
4182 Controls whether the memory cache is shared among SMP workers.
4184 The shared memory cache is meant to occupy cache_mem bytes and replace
4185 the non-shared memory cache, although some entities may still be
4186 cached locally by workers for now (e.g., internal and in-transit
4187 objects may be served from a local memory cache even if shared memory
4188 caching is enabled).
4190 By default, the memory cache is shared if and only if all of the
4191 following conditions are satisfied: Squid runs in SMP mode with
4192 multiple workers, cache_mem is positive, and Squid environment
4193 supports required IPC primitives (e.g., POSIX shared memory segments
4194 and GCC-style atomic operations).
4196 To avoid blocking locks, shared memory uses opportunistic algorithms
4197 that do not guarantee that every cachable entity that could have been
4198 shared among SMP workers will actually be shared.
4201 NAME: memory_cache_mode
4205 DEFAULT_DOC: Keep the most recently fetched objects in memory
4207 Controls which objects to keep in the memory cache (cache_mem)
4209 always Keep most recently fetched objects in memory (default)
4211 disk Only disk cache hits are kept in memory, which means
4212 an object must first be cached on disk and then hit
4213 a second time before cached in memory.
4215 network Only objects fetched from network is kept in memory
4218 NAME: memory_replacement_policy
4220 LOC: Config.memPolicy
4223 The memory replacement policy parameter determines which
4224 objects are purged from memory when memory space is needed.
4226 See cache_replacement_policy for details on algorithms.
4231 -----------------------------------------------------------------------------
4234 NAME: cache_replacement_policy
4236 LOC: Config.replPolicy
4239 The cache replacement policy parameter determines which
4240 objects are evicted (replaced) when disk space is needed.
4242 lru : Squid's original list based LRU policy
4243 heap GDSF : Greedy-Dual Size Frequency
4244 heap LFUDA: Least Frequently Used with Dynamic Aging
4245 heap LRU : LRU policy implemented using a heap
4247 Applies to any cache_dir lines listed below this directive.
4249 The LRU policies keeps recently referenced objects.
4251 The heap GDSF policy optimizes object hit rate by keeping smaller
4252 popular objects in cache so it has a better chance of getting a
4253 hit. It achieves a lower byte hit rate than LFUDA though since
4254 it evicts larger (possibly popular) objects.
4256 The heap LFUDA policy keeps popular objects in cache regardless of
4257 their size and thus optimizes byte hit rate at the expense of
4258 hit rate since one large, popular object will prevent many
4259 smaller, slightly less popular objects from being cached.
4261 Both policies utilize a dynamic aging mechanism that prevents
4262 cache pollution that can otherwise occur with frequency-based
4263 replacement policies.
4265 NOTE: if using the LFUDA replacement policy you should increase
4266 the value of maximum_object_size above its default of 4 MB to
4267 to maximize the potential byte hit rate improvement of LFUDA.
4269 For more information about the GDSF and LFUDA cache replacement
4270 policies see http://www.hpl.hp.com/techreports/1999/HPL-1999-69.html
4271 and http://fog.hpl.external.hp.com/techreports/98/HPL-98-173.html.
4274 NAME: minimum_object_size
4278 DEFAULT_DOC: no limit
4279 LOC: Config.Store.minObjectSize
4281 Objects smaller than this size will NOT be saved on disk. The
4282 value is specified in bytes, and the default is 0 KB, which
4283 means all responses can be stored.
4286 NAME: maximum_object_size
4290 LOC: Config.Store.maxObjectSize
4292 Set the default value for max-size parameter on any cache_dir.
4293 The value is specified in bytes, and the default is 4 MB.
4295 If you wish to get a high BYTES hit ratio, you should probably
4296 increase this (one 32 MB object hit counts for 3200 10KB
4299 If you wish to increase hit ratio more than you want to
4300 save bandwidth you should leave this low.
4302 NOTE: if using the LFUDA replacement policy you should increase
4303 this value to maximize the byte hit rate improvement of LFUDA!
4304 See cache_replacement_policy for a discussion of this policy.
4310 DEFAULT_DOC: No disk cache. Store cache objects only in memory.
4311 LOC: Config.cacheSwap
4314 cache_dir Type Directory-Name Fs-specific-data [options]
4316 You can specify multiple cache_dir lines to spread the
4317 cache among different disk partitions.
4319 Type specifies the kind of storage system to use. Only "ufs"
4320 is built by default. To enable any of the other storage systems
4321 see the --enable-storeio configure option.
4323 'Directory' is a top-level directory where cache swap
4324 files will be stored. If you want to use an entire disk
4325 for caching, this can be the mount-point directory.
4326 The directory must exist and be writable by the Squid
4327 process. Squid will NOT create this directory for you.
4329 Rock is currently the only SMP-aware cache_dir type. Using other
4330 store types in configurations with multiple workers is not
4331 supported and may lead to HTTP violations or undefined behavior,
4332 even when each such cache_dir is given a dedicated worker using
4333 configuration conditionals.
4336 ==== The ufs store type ====
4338 "ufs" is the old well-known Squid storage format that has always
4342 cache_dir ufs Directory-Name Mbytes L1 L2 [options]
4344 'Mbytes' is the amount of disk space (MB) to use under this
4345 directory. The default is 100 MB. Change this to suit your
4346 configuration. Do NOT put the size of your disk drive here.
4347 Instead, if you want Squid to use the entire disk drive,
4348 subtract 20% and use that value.
4350 'L1' is the number of first-level subdirectories which
4351 will be created under the 'Directory'. The default is 16.
4353 'L2' is the number of second-level subdirectories which
4354 will be created under each first-level directory. The default
4358 ==== The aufs store type ====
4360 "aufs" uses the same storage format as "ufs", utilizing
4361 POSIX-threads to avoid blocking the main Squid process on
4362 disk-I/O. This was formerly known in Squid as async-io.
4365 cache_dir aufs Directory-Name Mbytes L1 L2 [options]
4367 see argument descriptions under ufs above
4370 ==== The diskd store type ====
4372 "diskd" uses the same storage format as "ufs", utilizing a
4373 separate process to avoid blocking the main Squid process on
4377 cache_dir diskd Directory-Name Mbytes L1 L2 [options] [Q1=n] [Q2=n]
4379 see argument descriptions under ufs above
4381 Q1 specifies the number of unacknowledged I/O requests when Squid
4382 stops opening new files. If this many messages are in the queues,
4383 Squid won't open new files. Default is 64
4385 Q2 specifies the number of unacknowledged messages when Squid
4386 starts blocking. If this many messages are in the queues,
4387 Squid blocks until it receives some replies. Default is 72
4389 When Q1 < Q2 (the default), the cache directory is optimized
4390 for lower response time at the expense of a decrease in hit
4391 ratio. If Q1 > Q2, the cache directory is optimized for
4392 higher hit ratio at the expense of an increase in response
4396 ==== The rock store type ====
4399 cache_dir rock Directory-Name Mbytes [options]
4401 The Rock Store type is a database-style storage. All cached
4402 entries are stored in a "database" file, using fixed-size slots.
4403 A single entry occupies one or more slots.
4405 If possible, Squid using Rock Store creates a dedicated kid
4406 process called "disker" to avoid blocking Squid worker(s) on disk
4407 I/O. One disker kid is created for each rock cache_dir. Diskers
4408 are created only when Squid, running in daemon mode, has support
4409 for the IpcIo disk I/O module.
4411 swap-timeout=msec: Squid will not start writing a miss to or
4412 reading a hit from disk if it estimates that the swap operation
4413 will take more than the specified number of milliseconds. By
4414 default and when set to zero, disables the disk I/O time limit
4415 enforcement. Ignored when using blocking I/O module because
4416 blocking synchronous I/O does not allow Squid to estimate the
4417 expected swap wait time.
4419 max-swap-rate=swaps/sec: Artificially limits disk access using
4420 the specified I/O rate limit. Swap out requests that
4421 would cause the average I/O rate to exceed the limit are
4422 delayed. Individual swap in requests (i.e., hits or reads) are
4423 not delayed, but they do contribute to measured swap rate and
4424 since they are placed in the same FIFO queue as swap out
4425 requests, they may wait longer if max-swap-rate is smaller.
4426 This is necessary on file systems that buffer "too
4427 many" writes and then start blocking Squid and other processes
4428 while committing those writes to disk. Usually used together
4429 with swap-timeout to avoid excessive delays and queue overflows
4430 when disk demand exceeds available disk "bandwidth". By default
4431 and when set to zero, disables the disk I/O rate limit
4432 enforcement. Currently supported by IpcIo module only.
4434 slot-size=bytes: The size of a database "record" used for
4435 storing cached responses. A cached response occupies at least
4436 one slot and all database I/O is done using individual slots so
4437 increasing this parameter leads to more disk space waste while
4438 decreasing it leads to more disk I/O overheads. Should be a
4439 multiple of your operating system I/O page size. Defaults to
4440 16KBytes. A housekeeping header is stored with each slot and
4441 smaller slot-sizes will be rejected. The header is smaller than
4445 ==== COMMON OPTIONS ====
4447 no-store no new objects should be stored to this cache_dir.
4449 min-size=n the minimum object size in bytes this cache_dir
4450 will accept. It's used to restrict a cache_dir
4451 to only store large objects (e.g. AUFS) while
4452 other stores are optimized for smaller objects
4456 max-size=n the maximum object size in bytes this cache_dir
4458 The value in maximum_object_size directive sets
4459 the default unless more specific details are
4460 available (ie a small store capacity).
4462 Note: To make optimal use of the max-size limits you should order
4463 the cache_dir lines with the smallest max-size value first.
4467 # Uncomment and adjust the following to add a disk cache directory.
4468 #cache_dir ufs @DEFAULT_SWAP_DIR@ 100 16 256
4472 NAME: store_dir_select_algorithm
4474 LOC: Config.store_dir_select_algorithm
4477 How Squid selects which cache_dir to use when the response
4478 object will fit into more than one.
4480 Regardless of which algorithm is used the cache_dir min-size
4481 and max-size parameters are obeyed. As such they can affect
4482 the selection algorithm by limiting the set of considered
4489 This algorithm is suited to caches with similar cache_dir
4490 sizes and disk speeds.
4492 The disk with the least I/O pending is selected.
4493 When there are multiple disks with the same I/O load ranking
4494 the cache_dir with most available capacity is selected.
4496 When a mix of cache_dir sizes are configured the faster disks
4497 have a naturally lower I/O loading and larger disks have more
4498 capacity. So space used to store objects and data throughput
4499 may be very unbalanced towards larger disks.
4504 This algorithm is suited to caches with unequal cache_dir
4507 Each cache_dir is selected in a rotation. The next suitable
4510 Available cache_dir capacity is only considered in relation
4511 to whether the object will fit and meets the min-size and
4512 max-size parameters.
4514 Disk I/O loading is only considered to prevent overload on slow
4515 disks. This algorithm does not spread objects by size, so any
4516 I/O loading per-disk may appear very unbalanced and volatile.
4518 If several cache_dirs use similar min-size, max-size, or other
4519 limits to to reject certain responses, then do not group such
4520 cache_dir lines together, to avoid round-robin selection bias
4521 towards the first cache_dir after the group. Instead, interleave
4522 cache_dir lines from different groups. For example:
4524 store_dir_select_algorithm round-robin
4525 cache_dir rock /hdd1 ... min-size=100000
4526 cache_dir rock /ssd1 ... max-size=99999
4527 cache_dir rock /hdd2 ... min-size=100000
4528 cache_dir rock /ssd2 ... max-size=99999
4529 cache_dir rock /hdd3 ... min-size=100000
4530 cache_dir rock /ssd3 ... max-size=99999
4533 NAME: paranoid_hit_validation
4534 COMMENT: time-units-small
4535 TYPE: time_nanoseconds
4537 DEFAULT_DOC: validation disabled
4538 LOC: Config.paranoid_hit_validation
4540 Controls whether Squid should perform paranoid validation of cache entry
4541 metadata integrity every time a cache entry is hit. This low-level
4542 validation should always succeed. Each failed validation results in a
4543 cache miss, a BUG line reported to cache.log, and the invalid entry
4544 marked as unusable (and eventually purged from the cache).
4546 Squid can only validate shared cache memory and rock cache_dir entries.
4548 * Zero (default) value means that the validation is disabled.
4550 * Positive values enable validation:
4551 - values less than 1 day approximate the maximum time that Squid is allowed
4552 to spend validating a single cache hit.
4553 - values greater or equal to 1 day are considered as no limitation:
4554 in this case all checks will be performed, regardless of how much time
4557 Hits are usually stored using 16KB slots (for rock, the size is
4558 configurable via cache_dir slot-size). Larger hits require scanning more
4559 slots and, hence, take more time. When validation is enabled, at least one
4560 slot is always validated, regardless of the configured time limit.
4562 A worker process validating an entry cannot do anything else (i.e. the
4563 validation is blocking). The validation overhead is environment dependent,
4564 but developers have observed Squid spending 3-10 microseconds to check each
4565 slot of a Rock or shared memory hit entry. If Squid cuts validation short
4566 because it runs out of configured time, it treats the entry as valid.
4568 When hit validation is enabled, its statistics is included in Cache
4569 Manager mgr:counters, mgr:5min, and mgr:60min reports.
4572 NAME: max_open_disk_fds
4574 LOC: Config.max_open_disk_fds
4576 DEFAULT_DOC: no limit
4578 To avoid having disk as the I/O bottleneck Squid can optionally
4579 bypass the on-disk cache if more than this amount of disk file
4580 descriptors are open.
4582 A value of 0 indicates no limit.
4585 NAME: cache_swap_low
4586 COMMENT: (percent, 0-100)
4589 LOC: Config.Swap.lowWaterMark
4591 The low-water mark for AUFS/UFS/diskd cache object eviction by
4592 the cache_replacement_policy algorithm.
4594 Removal begins when the swap (disk) usage of a cache_dir is
4595 above this low-water mark and attempts to maintain utilization
4596 near the low-water mark.
4598 As swap utilization increases towards the high-water mark set
4599 by cache_swap_high object eviction becomes more aggressive.
4601 The value difference in percentages between low- and high-water
4602 marks represent an eviction rate of 300 objects per second and
4603 the rate continues to scale in aggressiveness by multiples of
4604 this above the high-water mark.
4606 Defaults are 90% and 95%. If you have a large cache, 5% could be
4607 hundreds of MB. If this is the case you may wish to set these
4608 numbers closer together.
4610 See also cache_swap_high and cache_replacement_policy
4613 NAME: cache_swap_high
4614 COMMENT: (percent, 0-100)
4617 LOC: Config.Swap.highWaterMark
4619 The high-water mark for AUFS/UFS/diskd cache object eviction by
4620 the cache_replacement_policy algorithm.
4622 Removal begins when the swap (disk) usage of a cache_dir is
4623 above the low-water mark set by cache_swap_low and attempts to
4624 maintain utilization near the low-water mark.
4626 As swap utilization increases towards this high-water mark object
4627 eviction becomes more aggressive.
4629 The value difference in percentages between low- and high-water
4630 marks represent an eviction rate of 300 objects per second and
4631 the rate continues to scale in aggressiveness by multiples of
4632 this above the high-water mark.
4634 Defaults are 90% and 95%. If you have a large cache, 5% could be
4635 hundreds of MB. If this is the case you may wish to set these
4636 numbers closer together.
4638 See also cache_swap_low and cache_replacement_policy
4643 -----------------------------------------------------------------------------
4650 DEFAULT_DOC: The format definitions squid, common, combined, referrer, useragent are built in.
4654 logformat <name> <format specification>
4656 Defines an access log format.
4658 The <format specification> is a string with embedded % format codes
4660 % format codes all follow the same basic structure where all
4661 components but the formatcode are optional and usually unnecessary,
4662 especially when dealing with common codes.
4664 % [encoding] [-] [[0]width] [{arg}] formatcode [{arg}]
4666 encoding escapes or otherwise protects "special" characters:
4668 " Quoted string encoding where quote(") and
4669 backslash(\) characters are \-escaped while
4670 CR, LF, and TAB characters are encoded as \r,
4671 \n, and \t two-character sequences.
4673 [ Custom Squid encoding where percent(%), square
4674 brackets([]), backslash(\) and characters with
4675 codes outside of [32,126] range are %-encoded.
4676 SP is not encoded. Used by log_mime_hdrs.
4678 # URL encoding (a.k.a. percent-encoding) where
4679 all URL unsafe and control characters (per RFC
4680 1738) are %-encoded.
4682 / Shell-like encoding where quote(") and
4683 backslash(\) characters are \-escaped while CR
4684 and LF characters are encoded as \r and \n
4685 two-character sequences. Values containing SP
4686 character(s) are surrounded by quotes(").
4688 ' Raw/as-is encoding with no escaping/quoting.
4690 Default encoding: When no explicit encoding is
4691 specified, each %code determines its own encoding.
4692 Most %codes use raw/as-is encoding, but some codes use
4693 a so called "pass-through URL encoding" where all URL
4694 unsafe and control characters (per RFC 1738) are
4695 %-encoded, but the percent character(%) is left as is.
4699 width minimum and/or maximum field width:
4700 [width_min][.width_max]
4701 When minimum starts with 0, the field is zero-padded.
4702 String values exceeding maximum width are truncated.
4704 {arg} argument such as header name etc. This field may be
4705 placed before or after the token, but not both at once.
4709 % a literal % character
4711 byte{value} Adds a single byte with the given value (e.g., %byte{10}
4712 adds an ASCII LF character a.k.a. "new line" or "\n"). The value
4713 parameter is required and must be a positive decimal integer not
4714 exceeding 255. Zero-valued bytes (i.e. ASCII NUL characters) are
4717 sn Unique sequence number per log line entry
4718 err_code The ID of an error response served by Squid or
4719 a similar internal error identifier.
4721 err_detail Additional err_code-dependent error information. Multiple
4722 details are separated by the plus sign ('+'). Admins should not
4723 rely on a particular detail listing order, the uniqueness of the
4724 entries, or individual detail text stability. All those properties
4725 depend on many unstable factors, including external libraries.
4727 note The annotation specified by the argument. Also
4728 logs the adaptation meta headers set by the
4729 adaptation_meta configuration parameter.
4730 If no argument given all annotations logged.
4731 The argument may include a separator to use with
4734 By default, multiple note values are separated with ","
4735 and multiple notes are separated with "\r\n".
4736 When logging named notes with %{name}note, the
4737 explicitly configured separator is used between note
4738 values. When logging all notes with %note, the
4739 explicitly configured separator is used between
4740 individual notes. There is currently no way to
4741 specify both value and notes separators when logging
4742 all notes with %note.
4743 master_xaction The master transaction identifier is an unsigned
4744 integer. These IDs are guaranteed to monotonically
4745 increase within a single worker process lifetime, with
4746 higher values corresponding to transactions that were
4747 accepted or initiated later. Due to current implementation
4748 deficiencies, some IDs are skipped (i.e. never logged).
4749 Concurrent workers and restarted workers use similar,
4750 overlapping sequences of master transaction IDs.
4752 Connection related format codes:
4754 >a Client source IP address
4756 >p Client source port
4757 >eui Client source EUI (MAC address, EUI-48 or EUI-64 identifier)
4758 >la Local IP address the client connected to
4759 >lp Local port number the client connected to
4760 >qos Client connection TOS/DSCP value set by Squid
4761 >nfmark Client connection netfilter packet MARK set by Squid
4763 transport::>connection_id Identifies a transport connection
4764 accepted by Squid (e.g., a connection carrying the
4765 logged HTTP request). Currently, Squid only supports
4766 TCP transport connections.
4768 The logged identifier is an unsigned integer. These
4769 IDs are guaranteed to monotonically increase within a
4770 single worker process lifetime, with higher values
4771 corresponding to connections that were accepted later.
4772 Many IDs are skipped (i.e. never logged). Concurrent
4773 workers and restarted workers use similar, partially
4774 overlapping sequences of IDs.
4776 la Local listening IP address the client connection was connected to.
4777 lp Local listening port number the client connection was connected to.
4779 <a Server IP address of the last server or peer connection
4780 <A Server FQDN or peer name
4781 <p Server port number of the last server or peer connection
4782 <la Local IP address of the last server or peer connection
4783 <lp Local port number of the last server or peer connection
4784 <qos Server connection TOS/DSCP value set by Squid
4785 <nfmark Server connection netfilter packet MARK set by Squid
4787 >handshake Raw client handshake
4788 Initial client bytes received by Squid on a newly
4789 accepted TCP connection or inside a just established
4790 CONNECT tunnel. Squid stops accumulating handshake
4791 bytes as soon as the handshake parser succeeds or
4792 fails (determining whether the client is using the
4795 For HTTP clients, the handshake is the request line.
4796 For TLS clients, the handshake consists of all TLS
4797 records up to and including the TLS record that
4798 contains the last byte of the first ClientHello
4799 message. For clients using an unsupported protocol,
4800 this field contains the bytes received by Squid at the
4801 time of the handshake parsing failure.
4803 See the on_unsupported_protocol directive for more
4804 information on Squid handshake traffic expectations.
4806 Current support is limited to these contexts:
4807 - http_port connections, but only when the
4808 on_unsupported_protocol directive is in use.
4809 - https_port connections (and CONNECT tunnels) that
4810 are subject to the ssl_bump peek or stare action.
4812 To protect binary handshake data, this field is always
4813 base64-encoded (RFC 4648 Section 4). If logformat
4814 field encoding is configured, that encoding is applied
4815 on top of base64. Otherwise, the computed base64 value
4818 Time related format codes:
4820 ts Seconds since epoch
4821 tu subsecond time (milliseconds)
4822 tl Local time. Optional strftime format argument
4823 default %d/%b/%Y:%H:%M:%S %z
4824 tg GMT time. Optional strftime format argument
4825 default %d/%b/%Y:%H:%M:%S %z
4826 tr Response time (milliseconds)
4827 dt Total time spent making DNS lookups (milliseconds)
4828 tS Approximate master transaction start time in
4829 <full seconds since epoch>.<fractional seconds> format.
4830 Currently, Squid considers the master transaction
4831 started when a complete HTTP request header initiating
4832 the transaction is received from the client. This is
4833 the same value that Squid uses to calculate transaction
4834 response time when logging %tr to access.log. Currently,
4835 Squid uses millisecond resolution for %tS values,
4836 similar to the default access.log "current time" field
4839 busy_time Time spent in transaction-related code (nanoseconds)
4840 This cumulative measurement excludes periods of time when the
4841 transaction was waiting (e.g., for a server or helper response)
4842 while Squid worked on other transactions or was engaged in
4843 transaction-unrelated activities (e.g., generating a cache index).
4844 In other words, this measurement represents the total amount of
4845 physical time when Squid was busy working on this transaction.
4847 WARNING: This measurement relies on Squid transaction context
4848 tracking features that currently have known context leak bugs and
4849 coverage gaps. Until those features are fully implemented, logged
4850 values may significantly understate or exaggerate actual times.
4851 Do not use this measurement unless you know it works in your case.
4853 Access Control related format codes:
4855 et Tag returned by external acl
4856 ea Log string returned by external acl
4857 un User name (any available)
4858 ul User name from authentication
4859 ue User name from external acl helper
4860 un A user name. Expands to the first available name
4861 from the following list of information sources:
4862 - authenticated user name, like %ul
4863 - user name supplied by an external ACL, like %ue
4864 - SSL client name, like %us
4865 credentials Client credentials. The exact meaning depends on
4866 the authentication scheme: For Basic authentication,
4867 it is the password; for Digest, the realm sent by the
4868 client; for NTLM and Negotiate, the client challenge
4869 or client credentials prefixed with "YR " or "KK ".
4871 HTTP related format codes:
4875 [http::]rm Request method (GET/POST etc)
4876 [http::]>rm Request method from client
4877 [http::]<rm Request method sent to server or peer
4879 [http::]ru Request URL received (or computed) and sanitized
4881 Logs request URI received from the client, a
4882 request adaptation service, or a request
4883 redirector (whichever was applied last).
4885 Computed URLs are URIs of internally generated
4886 requests and various "error:..." URIs.
4888 Honors strip_query_terms and uri_whitespace.
4890 This field is not encoded by default. Encoding
4891 this field using variants of %-encoding will
4892 clash with uri_whitespace modifications that
4893 also use %-encoding.
4895 [http::]>ru Request URL received from the client (or computed)
4897 Computed URLs are URIs of internally generated
4898 requests and various "error:..." URIs.
4900 Unlike %ru, this request URI is not affected
4901 by request adaptation, URL rewriting services,
4902 and strip_query_terms.
4904 Honors uri_whitespace.
4906 This field is using pass-through URL encoding
4907 by default. Encoding this field using other
4908 variants of %-encoding will clash with
4909 uri_whitespace modifications that also use
4912 [http::]<ru Request URL sent to server or peer
4913 [http::]>rs Request URL scheme from client
4914 [http::]<rs Request URL scheme sent to server or peer
4915 [http::]>rd Request URL domain from client
4916 [http::]<rd Request URL domain sent to server or peer
4917 [http::]>rP Request URL port from client
4918 [http::]<rP Request URL port sent to server or peer
4919 [http::]rp Request URL path excluding hostname
4920 [http::]>rp Request URL path excluding hostname from client
4921 [http::]<rp Request URL path excluding hostname sent to server or peer
4922 [http::]rv Request protocol version
4923 [http::]>rv Request protocol version from client
4924 [http::]<rv Request protocol version sent to server or peer
4926 [http::]>h Original received request header.
4927 Usually differs from the request header sent by
4928 Squid, although most fields are often preserved.
4929 Accepts optional header field name/value filter
4930 argument using name[:[separator]element] format.
4931 [http::]>ha Received request header after adaptation and
4932 redirection (pre-cache REQMOD vectoring point).
4933 Usually differs from the request header sent by
4934 Squid, although most fields are often preserved.
4935 Optional header name argument as for >h
4939 [http::]<Hs HTTP status code received from the next hop
4940 [http::]>Hs HTTP status code sent to the client
4942 [http::]<h Reply header. Optional header name argument
4945 [http::]mt MIME content type
4950 [http::]st Total size of request + reply traffic with client
4951 [http::]>st Total size of request received from client.
4952 Excluding chunked encoding bytes.
4953 [http::]<st Total size of reply sent to client (after adaptation)
4955 [http::]>sh Size of request headers received from client
4956 [http::]<sh Size of reply headers sent to client (after adaptation)
4958 [http::]<sH Reply high offset sent
4959 [http::]<sS Upstream object size
4961 [http::]<bs Number of HTTP-equivalent message body bytes
4962 received from the next hop, excluding chunked
4963 transfer encoding and control messages.
4964 Generated FTP listings are treated as
4969 [http::]<pt Peer response time in milliseconds. The timer starts
4970 when the last request byte is sent to the next hop
4971 and stops when the last response byte is received.
4972 [http::]<tt Total time spent forwarding to origin servers or
4973 cache_peers (milliseconds).
4975 The timer starts when Squid decides to forward the request (to
4976 an origin server or cache_peer) and peer selection begins. The
4977 timer stops when relevant forwarding activities (including any
4980 Between those two timer events, Squid may perform DNS lookups,
4981 query external ACL helpers, adapt responses using pre-cache
4982 RESPMOD services, and participate in other concurrent
4983 secondary activities. Most secondary activities increase
4984 peering time. In some cases, a secondary activity may start
4985 before the timer starts or end after the timer stops, leading
4986 to misleading results of simple computations like %<tt - %dt.
4988 If this logformat %code is used before its timer starts, the
4989 corresponding measurement has no value (and the %code expands
4990 to a single dash ("-") character).
4992 If this code is used while its timer is running, the time
4993 spent so far is used as the measurement value.
4995 When Squid re-forwards the request (e.g., after certain cache
4996 revalidation failures), the timer may restart. In this case,
4997 the new measurement is added to the value accumulated from
4998 previous forwarding attempts. The time interval between
4999 forwarding attempts is not added to the final result.
5001 Squid handling related format codes:
5003 Ss Squid request status (TCP_MISS etc)
5004 Sh Squid hierarchy status (DEFAULT_PARENT etc)
5006 [http::]request_attempts Number of request forwarding attempts
5008 See forward_max_tries documentation that details what Squid counts
5009 as a forwarding attempt. Pure cache hits log zero, but cache hits
5010 that triggered HTTP cache revalidation log the number of attempts
5011 made when sending an internal revalidation request. DNS, ICMP,
5012 ICP, HTCP, ICAP, eCAP, helper, and other secondary requests
5013 sent by Squid as a part of a master transaction do not increment
5014 the counter logged for the received request.
5016 SSL-related format codes:
5018 ssl::bump_mode SslBump decision for the transaction:
5020 For CONNECT requests that initiated bumping of
5021 a connection and for any request received on
5022 an already bumped connection, Squid logs the
5023 corresponding SslBump mode ("splice", "bump",
5024 "peek", "stare", "terminate", "server-first"
5025 or "client-first"). See the ssl_bump option
5026 for more information about these modes.
5028 A "none" token is logged for requests that
5029 triggered "ssl_bump" ACL evaluation matching
5032 In all other cases, a single dash ("-") is
5035 ssl::>sni SSL client SNI sent to Squid.
5038 The Subject field of the received client
5039 SSL certificate or a dash ('-') if Squid has
5040 received an invalid/malformed certificate or
5041 no certificate at all. Consider encoding the
5042 logged value because Subject often has spaces.
5045 The Issuer field of the received client
5046 SSL certificate or a dash ('-') if Squid has
5047 received an invalid/malformed certificate or
5048 no certificate at all. Consider encoding the
5049 logged value because Issuer often has spaces.
5052 The Subject field of the received server
5053 TLS certificate or a dash ('-') if this is
5054 not available. Consider encoding the logged
5055 value because Subject often has spaces.
5058 The Issuer field of the received server
5059 TLS certificate or a dash ('-') if this is
5060 not available. Consider encoding the logged
5061 value because Issuer often has spaces.
5064 The received server x509 certificate in PEM
5065 format, including BEGIN and END lines (or a
5066 dash ('-') if the certificate is unavailable).
5068 WARNING: Large certificates will exceed the
5069 current 8KB access.log record limit, resulting
5070 in truncated records. Such truncation usually
5071 happens in the middle of a record field. The
5072 limit applies to all access logging modules.
5074 The logged certificate may have failed
5075 validation and may not be trusted by Squid.
5076 This field does not include any intermediate
5077 certificates that may have been received from
5078 the server or fetched during certificate
5081 Currently, Squid only collects server
5082 certificates during step3 of SslBump
5083 processing; connections that were not subject
5084 to ssl_bump rules or that did not match a peek
5085 or stare rule at step2 will not have the
5086 server certificate information.
5088 This field is using pass-through URL encoding
5092 The list of certificate validation errors
5093 detected by Squid (including OpenSSL and
5094 certificate validation helper components). The
5095 errors are listed in the discovery order. By
5096 default, the error codes are separated by ':'.
5097 Accepts an optional separator argument.
5099 %ssl::>negotiated_version The negotiated TLS version of the
5102 %ssl::<negotiated_version The negotiated TLS version of the
5103 last server or peer connection.
5105 %ssl::>received_hello_version The TLS version of the Hello
5106 message received from TLS client.
5108 %ssl::<received_hello_version The TLS version of the Hello
5109 message received from TLS server.
5111 %ssl::>received_supported_version The maximum TLS version
5112 supported by the TLS client.
5114 %ssl::<received_supported_version The maximum TLS version
5115 supported by the TLS server.
5117 %ssl::>negotiated_cipher The negotiated cipher of the
5120 %ssl::<negotiated_cipher The negotiated cipher of the
5121 last server or peer connection.
5123 If ICAP is enabled, the following code becomes available (as
5124 well as ICAP log codes documented with the icap_log option):
5126 icap::tt Total ICAP "blocking" time for the HTTP transaction. The
5127 timer ticks while Squid checks adaptation_access and while
5128 ICAP transaction(s) expect ICAP response headers, including
5129 the embedded adapted HTTP message headers (where applicable).
5130 This measurement is meant to estimate ICAP impact on HTTP
5131 transaction response times, but it does not currently account
5132 for slow ICAP response body delivery blocking HTTP progress.
5134 Once Squid receives the final ICAP response headers (e.g.,
5135 ICAP 200 or 204) and the associated adapted HTTP message
5136 headers (if any) from the ICAP service, the corresponding ICAP
5137 transaction stops affecting this measurement, even though the
5138 transaction itself may continue for a long time (e.g., to
5139 finish sending the ICAP request and/or to finish receiving the
5140 ICAP response body).
5142 When "blocking" sections of multiple concurrent ICAP
5143 transactions overlap in time, the overlapping segment is
5146 To see complete ICAP transaction response times (rather than
5147 the cumulative effect of their blocking sections) use the
5148 %adapt::all_trs logformat code or the icap_log directive.
5150 If adaptation is enabled the following codes become available:
5152 adapt::<last_h The header of the last ICAP response or
5153 meta-information from the last eCAP
5154 transaction related to the HTTP transaction.
5155 Like <h, accepts an optional header name
5158 adapt::sum_trs Summed adaptation transaction response
5159 times recorded as a comma-separated list in
5160 the order of transaction start time. Each time
5161 value is recorded as an integer number,
5162 representing response time of one or more
5163 adaptation (ICAP or eCAP) transaction in
5164 milliseconds. When a failed transaction is
5165 being retried or repeated, its time is not
5166 logged individually but added to the
5167 replacement (next) transaction. Lifetimes of individually
5168 listed adaptation transactions may overlap.
5169 See also: %icap::tt and %adapt::all_trs.
5171 adapt::all_trs All adaptation transaction response times.
5172 Same as %adapt::sum_trs but response times of
5173 individual transactions are never added
5174 together. Instead, all transaction response
5175 times are recorded individually.
5177 You can prefix adapt::*_trs format codes with adaptation
5178 service name in curly braces to record response time(s) specific
5179 to that service. For example: %{my_service}adapt::sum_trs
5181 Format codes related to the PROXY protocol:
5183 proxy_protocol::>h PROXY protocol header, including optional TLVs.
5185 Supports the same field and element reporting/extraction logic
5186 as %http::>h. For configuration and reporting purposes, Squid
5187 maps each PROXY TLV to an HTTP header field: the TLV type
5188 (configured as a decimal integer) is the field name, and the
5189 TLV value is the field value. All TLVs of "LOCAL" connections
5190 (in PROXY protocol terminology) are currently skipped/ignored.
5192 Squid also maps the following standard PROXY protocol header
5193 blocks to pseudo HTTP headers (their names use PROXY
5194 terminology and start with a colon, following HTTP tradition
5195 for pseudo headers): :command, :version, :src_addr, :dst_addr,
5196 :src_port, and :dst_port.
5198 Without optional parameters, this logformat code logs
5199 pseudo headers and TLVs.
5201 This format code uses pass-through URL encoding by default.
5204 # relay custom PROXY TLV #224 to adaptation services
5205 adaptation_meta Client-Foo "%proxy_protocol::>h{224}"
5209 The default formats available (which do not need re-defining) are:
5211 logformat squid %ts.%03tu %6tr %>a %Ss/%03>Hs %<st %rm %ru %[un %Sh/%<a %mt
5212 logformat common %>a - %[un [%tl] "%rm %ru HTTP/%rv" %>Hs %<st %Ss:%Sh
5213 logformat combined %>a - %[un [%tl] "%rm %ru HTTP/%rv" %>Hs %<st "%{Referer}>h" "%{User-Agent}>h" %Ss:%Sh
5214 logformat referrer %ts.%03tu %>a %{Referer}>h %ru
5215 logformat useragent %>a [%tl] "%{User-Agent}>h"
5217 NOTE: When the log_mime_hdrs directive is set to ON.
5218 The squid, common and combined formats have a safely encoded copy
5219 of the mime headers appended to each line within a pair of brackets.
5221 NOTE: The common and combined formats are not quite true to the Apache definition.
5222 The logs from Squid contain an extra status and hierarchy code appended.
5226 NAME: access_log cache_access_log
5228 LOC: Config.Log.accesslogs
5229 DEFAULT_IF_NONE: daemon:@DEFAULT_ACCESS_LOG@ squid
5231 Configures whether and how Squid logs HTTP and ICP transactions.
5232 If access logging is enabled, a single line is logged for every
5233 matching HTTP or ICP request. The recommended directive formats are:
5235 access_log <module>:<place> [option ...] [acl acl ...]
5236 access_log none [acl acl ...]
5238 The following directive format is accepted but may be deprecated:
5239 access_log <module>:<place> [<logformat name> [acl acl ...]]
5241 In most cases, the first ACL name must not contain the '=' character
5242 and should not be equal to an existing logformat name. You can always
5243 start with an 'all' ACL to work around those restrictions.
5245 Will log to the specified module:place using the specified format (which
5246 must be defined in a logformat directive) those entries which match
5247 ALL the acl's specified (which must be defined in acl clauses).
5248 If no acl is specified, all requests will be logged to this destination.
5250 ===== Available options for the recommended directive format =====
5252 logformat=name Names log line format (either built-in or
5253 defined by a logformat directive). Defaults
5256 buffer-size=64KB Defines approximate buffering limit for log
5257 records (see buffered_logs). Squid should not
5258 keep more than the specified size and, hence,
5259 should flush records before the buffer becomes
5260 full to avoid overflows under normal
5261 conditions (the exact flushing algorithm is
5262 module-dependent though). The on-error option
5263 controls overflow handling.
5265 on-error=die|drop Defines action on unrecoverable errors. The
5266 'drop' action ignores (i.e., does not log)
5267 affected log records. The default 'die' action
5268 kills the affected worker. The drop action
5269 support has not been tested for modules other
5272 rotate=N Specifies the number of log file rotations to
5273 make when you run 'squid -k rotate'. The default
5274 is to obey the logfile_rotate directive. Setting
5275 rotate=0 will disable the file name rotation,
5276 but the log files are still closed and re-opened.
5277 This will enable you to rename the logfiles
5278 yourself just before sending the rotate signal.
5279 Only supported by the stdio module.
5281 ===== Modules Currently available =====
5283 none Do not log any requests matching these ACL.
5284 Do not specify Place or logformat name.
5286 stdio Write each log line to disk immediately at the completion of
5288 Place: the filename and path to be written.
5290 daemon Very similar to stdio. But instead of writing to disk the log
5291 line is passed to a daemon helper for asychronous handling instead.
5292 Place: varies depending on the daemon.
5294 log_file_daemon Place: the file name and path to be written.
5296 syslog To log each request via syslog facility.
5297 Place: The syslog facility and priority level for these entries.
5298 Place Format: facility.priority
5300 where facility could be any of:
5301 authpriv, daemon, local0 ... local7 or user.
5303 And priority could be any of:
5304 err, warning, notice, info, debug.
5306 udp To send each log line as text data to a UDP receiver.
5307 Place: The destination host name or IP and port.
5308 Place Format: //host:port
5310 tcp To send each log line as text data to a TCP receiver.
5311 Lines may be accumulated before sending (see buffered_logs).
5312 Place: The destination host name or IP and port.
5313 Place Format: //host:port
5316 access_log daemon:@DEFAULT_ACCESS_LOG@ squid
5322 LOC: Config.Log.icaplogs
5325 ICAP log files record ICAP transaction summaries, one line per
5328 The icap_log option format is:
5329 icap_log <filepath> [<logformat name> [acl acl ...]]
5330 icap_log none [acl acl ...]]
5332 Please see access_log option documentation for details. The two
5333 kinds of logs share the overall configuration approach and many
5336 ICAP processing of a single HTTP message or transaction may
5337 require multiple ICAP transactions. In such cases, multiple
5338 ICAP transaction log lines will correspond to a single access
5341 ICAP log supports many access.log logformat %codes. In ICAP context,
5342 HTTP message-related %codes are applied to the HTTP message embedded
5343 in an ICAP message. Logformat "%http::>..." codes are used for HTTP
5344 messages embedded in ICAP requests while "%http::<..." codes are used
5345 for HTTP messages embedded in ICAP responses. For example:
5347 http::>h To-be-adapted HTTP message headers sent by Squid to
5348 the ICAP service. For REQMOD transactions, these are
5349 HTTP request headers. For RESPMOD, these are HTTP
5350 response headers, but Squid currently cannot log them
5351 (i.e., %http::>h will expand to "-" for RESPMOD).
5353 http::<h Adapted HTTP message headers sent by the ICAP
5354 service to Squid (i.e., HTTP request headers in regular
5355 REQMOD; HTTP response headers in RESPMOD and during
5356 request satisfaction in REQMOD).
5358 ICAP OPTIONS transactions do not embed HTTP messages.
5360 Several logformat codes below deal with ICAP message bodies. An ICAP
5361 message body, if any, typically includes a complete HTTP message
5362 (required HTTP headers plus optional HTTP message body). When
5363 computing HTTP message body size for these logformat codes, Squid
5364 either includes or excludes chunked encoding overheads; see
5365 code-specific documentation for details.
5367 For Secure ICAP services, all size-related information is currently
5368 computed before/after TLS encryption/decryption, as if TLS was not
5371 The following format codes are also available for ICAP logs:
5373 icap::<A ICAP server IP address. Similar to <A.
5375 icap::<service_name ICAP service name from the icap_service
5376 option in Squid configuration file.
5378 icap::ru ICAP Request-URI. Similar to ru.
5380 icap::rm ICAP request method (REQMOD, RESPMOD, or
5381 OPTIONS). Similar to existing rm.
5383 icap::>st The total size of the ICAP request sent to the ICAP
5384 server (ICAP headers + ICAP body), including chunking
5387 icap::<st The total size of the ICAP response received from the
5388 ICAP server (ICAP headers + ICAP body), including
5389 chunking metadata (if any).
5391 icap::<bs The size of the ICAP response body received from the
5392 ICAP server, excluding chunking metadata (if any).
5394 icap::tr Transaction response time (in
5395 milliseconds). The timer starts when
5396 the ICAP transaction is created and
5397 stops when the transaction is completed.
5400 icap::tio Transaction I/O time (in milliseconds). The
5401 timer starts when the first ICAP request
5402 byte is scheduled for sending. The timers
5403 stops when the last byte of the ICAP response
5406 icap::to Transaction outcome: ICAP_ERR* for all
5407 transaction errors, ICAP_OPT for OPTION
5408 transactions, ICAP_ECHO for 204
5409 responses, ICAP_MOD for message
5410 modification, and ICAP_SAT for request
5411 satisfaction. Similar to Ss.
5413 icap::Hs ICAP response status code. Similar to Hs.
5415 icap::>h ICAP request header(s). Similar to >h.
5417 icap::<h ICAP response header(s). Similar to <h.
5419 The default ICAP log format, which can be used without an explicit
5420 definition, is called icap_squid:
5422 logformat icap_squid %ts.%03tu %6icap::tr %>A %icap::to/%03icap::Hs %icap::<st %icap::rm %icap::ru %un -/%icap::<A -
5424 See also: logformat and %adapt::<last_h
5427 NAME: logfile_daemon
5429 DEFAULT: @DEFAULT_LOGFILED@
5430 LOC: Log::TheConfig.logfile_daemon
5432 Specify the path to the logfile-writing daemon. This daemon is
5433 used to write the access and store logs, if configured.
5435 Squid sends a number of commands to the log daemon:
5436 L<data>\n - logfile data
5441 r<n>\n - set rotate count to <n>
5442 b<n>\n - 1 = buffer output, 0 = don't buffer output
5444 No responses is expected.
5447 NAME: stats_collection
5449 LOC: Config.accessList.stats_collection
5451 DEFAULT_DOC: Allow logging for all transactions.
5452 COMMENT: allow|deny acl acl...
5454 This options allows you to control which requests gets accounted
5455 in performance counters.
5457 This clause only supports fast acl types.
5458 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
5461 NAME: cache_store_log
5464 LOC: Config.Log.store
5466 Logs the activities of the storage manager. Shows which
5467 objects are ejected from the cache, and which objects are
5468 saved and for how long.
5469 There are not really utilities to analyze this data, so you can safely
5470 disable it (the default).
5472 Store log uses modular logging outputs. See access_log for the list
5473 of modules supported.
5476 cache_store_log stdio:@DEFAULT_STORE_LOG@
5477 cache_store_log daemon:@DEFAULT_STORE_LOG@
5480 NAME: cache_swap_state cache_swap_log
5482 LOC: Config.Log.swap
5484 DEFAULT_DOC: Store the journal inside its cache_dir
5486 Location for the cache "swap.state" file. This index file holds
5487 the metadata of objects saved on disk. It is used to rebuild
5488 the cache during startup. Normally this file resides in each
5489 'cache_dir' directory, but you may specify an alternate
5490 pathname here. Note you must give a full filename, not just
5491 a directory. Since this is the index for the whole object
5492 list you CANNOT periodically rotate it!
5494 If %s can be used in the file name it will be replaced with a
5495 a representation of the cache_dir name where each / is replaced
5496 with '.'. This is needed to allow adding/removing cache_dir
5497 lines when cache_swap_log is being used.
5499 If have more than one 'cache_dir', and %s is not used in the name
5500 these swap logs will have names such as:
5506 The numbered extension (which is added automatically)
5507 corresponds to the order of the 'cache_dir' lines in this
5508 configuration file. If you change the order of the 'cache_dir'
5509 lines in this file, these index files will NOT correspond to
5510 the correct 'cache_dir' entry (unless you manually rename
5511 them). We recommend you do NOT use this option. It is
5512 better to keep these index files in each 'cache_dir' directory.
5515 NAME: logfile_rotate
5518 LOC: Config.Log.rotateNumber
5520 Specifies the default number of logfile rotations to make when you
5521 type 'squid -k rotate'. The default is 10, which will rotate
5522 with extensions 0 through 9. Setting logfile_rotate to 0 will
5523 disable the file name rotation, but the logfiles are still closed
5524 and re-opened. This will enable you to rename the logfiles
5525 yourself just before sending the rotate signal.
5527 Note, from Squid-3.1 this option is only a default for cache.log,
5528 that log can be rotated separately by using debug_options.
5530 Note, from Squid-4 this option is only a default for access.log
5531 recorded by stdio: module. Those logs can be rotated separately by
5532 using the rotate=N option on their access_log directive.
5534 Note, the 'squid -k rotate' command normally sends a USR1
5535 signal to the running squid process. In certain situations
5536 (e.g. on Linux with Async I/O), USR1 is used for other
5537 purposes, so -k rotate uses another signal. It is best to get
5538 in the habit of using 'squid -k rotate' instead of 'kill -USR1
5545 DEFAULT: @DEFAULT_MIME_TABLE@
5546 LOC: Config.mimeTablePathname
5548 Path to Squid's icon configuration file.
5550 You shouldn't need to change this, but the default file contains
5551 examples and formatting information if you do.
5557 LOC: Config.onoff.log_mime_hdrs
5560 The Cache can record both the request and the response MIME
5561 headers for each HTTP transaction. The headers are encoded
5562 safely and will appear as two bracketed fields at the end of
5563 the access log (for either the native or httpd-emulated log
5564 formats). To enable this logging set log_mime_hdrs to 'on'.
5569 DEFAULT: @DEFAULT_PID_FILE@
5570 LOC: Config.pidFilename
5572 A filename to write the process-id to. To disable, enter "none".
5575 NAME: client_netmask
5577 LOC: Config.Addrs.client_netmask
5579 DEFAULT_DOC: Log full client IP address
5581 A netmask for client addresses in logfiles and cachemgr output.
5582 Change this to protect the privacy of your cache clients.
5583 A netmask of 255.255.255.0 will log all IP's in that range with
5584 the last digit set to '0'.
5587 NAME: strip_query_terms
5589 LOC: Config.onoff.strip_query_terms
5592 By default, Squid strips query terms from requested URLs before
5593 logging. This protects your user's privacy and reduces log size.
5595 When investigating HIT/MISS or other caching behaviour you
5596 will need to disable this to see the full URL used by Squid.
5603 LOC: Config.onoff.buffered_logs
5605 Whether to write/send access_log records ASAP or accumulate them and
5606 then write/send them in larger chunks. Buffering may improve
5607 performance because it decreases the number of I/Os. However,
5608 buffering increases the delay before log records become available to
5609 the final recipient (e.g., a disk file or logging daemon) and,
5610 hence, increases the risk of log records loss.
5612 Note that even when buffered_logs are off, Squid may have to buffer
5613 records if it cannot write/send them immediately due to pending I/Os
5614 (e.g., the I/O writing the previous log record) or connectivity loss.
5616 Currently honored by 'daemon', 'tcp' and 'udp' access_log modules only.
5619 NAME: netdb_filename
5621 DEFAULT: stdio:@DEFAULT_NETDB_FILE@
5622 LOC: Config.netdbFilename
5625 Where Squid stores it's netdb journal.
5626 When enabled this journal preserves netdb state between restarts.
5628 To disable, enter "none".
5632 TYPE: Security::KeyLog*
5634 LOC: Config.Log.tlsKeys
5637 Configures whether and where Squid records pre-master secret and
5638 related encryption details for TLS connections accepted or established
5639 by Squid. These connections include connections accepted at
5640 https_port, TLS connections opened to origin servers/cache_peers/ICAP
5641 services, and TLS tunnels bumped by Squid using the SslBump feature.
5642 This log (a.k.a. SSLKEYLOGFILE) is meant for triage with traffic
5643 inspection tools like Wireshark.
5645 tls_key_log <destination> [options] [if [!]<acl>...]
5647 WARNING: This log allows anybody to decrypt the corresponding
5648 encrypted TLS connections, both in-flight and postmortem.
5650 At most one log file is supported at this time. Repeated tls_key_log
5651 directives are treated as fatal configuration errors. By default, no
5652 log is created or updated.
5654 If the log file does not exist, Squid creates it. Otherwise, Squid
5655 appends an existing log file.
5657 The directive is consulted whenever a TLS connection is accepted or
5658 established by Squid. TLS connections that fail the handshake may be
5659 logged if Squid got enough information to form a log record. A record
5660 is logged only if all of the configured ACLs match.
5662 While transport-related ACLs like src and dst should work, Squid may
5663 not have access to higher-level information. For example, when logging
5664 accepted https_port connections, Squid does not yet have access to the
5665 expected HTTPS request. Similarly, an HTTPS response is not available
5666 when logging most TLS connections established by Squid.
5668 The log record format is meant to be compatible with TLS deciphering
5669 features of Wireshark which relies on fields like CLIENT_RANDOM and
5670 RSA Master-Key. A single log record usually spans multiple lines.
5671 Technical documentation for that format is maintained inside the
5672 Wireshark code (e.g., see tls_keylog_process_lines() comments as of
5673 Wireshark commit e3d44136f0f0026c5e893fa249f458073f3b7328). TLS key
5674 log does not support custom record formats.
5676 This clause only supports fast acl types.
5677 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
5679 See access_log's <module>:<place> parameter for a list of supported
5680 logging destinations.
5682 TLS key log supports all access_log key=value options with the
5683 exception of logformat=name.
5685 Requires Squid built with OpenSSL support.
5690 OPTIONS FOR TROUBLESHOOTING
5691 -----------------------------------------------------------------------------
5696 DEFAULT_IF_NONE: @DEFAULT_CACHE_LOG@
5697 LOC: Debug::cache_log
5699 Squid administrative logging file.
5701 This is where general information about Squid behavior goes. You can
5702 increase the amount of data logged to this file and how often it is
5703 rotated with "debug_options"
5706 NAME: cache_log_message
5707 TYPE: cache_log_message
5709 DEFAULT_DOC: Use debug_options.
5710 LOC: DebugMessagesConfig
5712 Configures logging of individual cache.log messages.
5714 cache_log_message id=<number> option...
5715 cache_log_message ids=<number>-<number> option...
5717 Most messages have _not_ been instrumented to support this directive
5718 yet. For the list of instrumented messages and their IDs, please see
5719 the doc/debug-messages.dox file.
5721 Message ID corresponds to the message semantics rather than message
5722 text or source code location. The ID is stable across Squid
5723 instances and versions. Substantial changes in message semantics
5724 result in a new ID assignment. To reduce the danger of suppressing
5725 an important log message, the old IDs of removed (or substantially
5726 changed) messages are never reused.
5728 If more than one cache_log_message directive refers to the same
5729 message ID, the last directive wins.
5731 Use ids=min-max syntax to apply the same message configuration to an
5732 inclusive range of message IDs. An ID range with N values has
5733 exactly the same effect as typing N cache_log_message lines.
5735 At least one option is required. Supported options are:
5737 level=<number>: The logging level to use for the message. Squid
5738 command line options (-s and -d) as well as the debug_options
5739 directive control which levels go to syslog, stderr, and/or
5740 cache.log. In most environments, using level=2 or higher stops
5741 Squid from logging the message anywhere. By default, the
5742 hard-coded message-specific level is used.
5744 limit=<number>: After logging the specified number of messages at
5745 the configured (or default) debugging level DL, start using
5746 level 3 (for DL 0 and 1) or 8 (for higher DL values). Usually,
5747 level-3+ messages are not logged anywhere so this option can
5748 often be used to effectively suppress the message. Each SMP
5749 Squid process gets the same limit.
5755 DEFAULT_DOC: Log all critical and important messages.
5756 LOC: Debug::debugOptions
5758 Logging options are set as section,level where each source file
5759 is assigned a unique section. Lower levels result in less
5760 output, Full debugging (level 9) can result in a very large
5761 log file, so be careful.
5763 The magic word "ALL" sets debugging levels for all sections.
5764 The default is to run with "ALL,1" to record important warnings.
5766 The rotate=N option can be used to keep more or less of these logs
5767 than would otherwise be kept by logfile_rotate.
5768 For most uses a single log should be enough to monitor current
5769 events affecting Squid.
5774 LOC: Config.coredump_dir
5775 DEFAULT_IF_NONE: none
5776 DEFAULT_DOC: Use the directory from where Squid was started.
5778 By default Squid leaves core files in the directory from where
5779 it was started. If you set 'coredump_dir' to a directory
5780 that exists, Squid will chdir() to that directory at startup
5781 and coredump files will be left there.
5783 In addition to changing the directory, the process permissions are updated
5784 to enable process tracing and/or coredump file generation. The details are
5785 OS-specific, but look for prctl(2) PR_SET_DUMPABLE and procctl(2)
5786 PROC_TRACE_CTL documentation as guiding examples.
5790 # Leave coredumps in the first cache dir
5791 coredump_dir @DEFAULT_SWAP_DIR@
5797 OPTIONS FOR FTP GATEWAYING
5798 -----------------------------------------------------------------------------
5804 LOC: Config.Ftp.anon_user
5806 If you want the anonymous login password to be more informative
5807 (and enable the use of picky FTP servers), set this to something
5808 reasonable for your domain, like wwwuser@somewhere.net
5810 The reason why this is domainless by default is the
5811 request can be made on the behalf of a user in any domain,
5812 depending on how the cache is used.
5813 Some FTP server also validate the email address is valid
5814 (for example perl.com).
5820 LOC: Config.Ftp.passive
5822 If your firewall does not allow Squid to use passive
5823 connections, turn off this option.
5825 Use of ftp_epsv_all option requires this to be ON.
5831 LOC: Config.Ftp.epsv_all
5833 FTP Protocol extensions permit the use of a special "EPSV ALL" command.
5835 NATs may be able to put the connection on a "fast path" through the
5836 translator, as the EPRT command will never be used and therefore,
5837 translation of the data portion of the segments will never be needed.
5839 When a client only expects to do two-way FTP transfers this may be
5841 If squid finds that it must do a three-way FTP transfer after issuing
5842 an EPSV ALL command, the FTP session will fail.
5844 If you have any doubts about this option do not use it.
5845 Squid will nicely attempt all other connection methods.
5847 Requires ftp_passive to be ON (default) for any effect.
5853 LOC: Config.accessList.ftp_epsv
5855 FTP Protocol extensions permit the use of a special "EPSV" command.
5857 NATs may be able to put the connection on a "fast path" through the
5858 translator using EPSV, as the EPRT command will never be used
5859 and therefore, translation of the data portion of the segments
5860 will never be needed.
5862 EPSV is often required to interoperate with FTP servers on IPv6
5863 networks. On the other hand, it may break some IPv4 servers.
5865 By default, EPSV may try EPSV with any FTP server. To fine tune
5866 that decision, you may restrict EPSV to certain clients or servers
5869 ftp_epsv allow|deny al1 acl2 ...
5871 WARNING: Disabling EPSV may cause problems with external NAT and IPv6.
5873 Only fast ACLs are supported.
5874 Requires ftp_passive to be ON (default) for any effect.
5880 LOC: Config.Ftp.eprt
5882 FTP Protocol extensions permit the use of a special "EPRT" command.
5884 This extension provides a protocol neutral alternative to the
5885 IPv4-only PORT command. When supported it enables active FTP data
5886 channels over IPv6 and efficient NAT handling.
5888 Turning this OFF will prevent EPRT being attempted and will skip
5889 straight to using PORT for IPv4 servers.
5891 Some devices are known to not handle this extension correctly and
5892 may result in crashes. Devices which support EPRT enough to fail
5893 cleanly will result in Squid attempting PORT anyway. This directive
5894 should only be disabled when EPRT results in device failures.
5896 WARNING: Doing so will convert Squid back to the old behavior with all
5897 the related problems with external NAT devices/layers and IPv4-only FTP.
5900 NAME: ftp_sanitycheck
5903 LOC: Config.Ftp.sanitycheck
5905 For security and data integrity reasons Squid by default performs
5906 sanity checks of the addresses of FTP data connections ensure the
5907 data connection is to the requested server. If you need to allow
5908 FTP connections to servers using another IP address for the data
5909 connection turn this off.
5912 NAME: ftp_telnet_protocol
5915 LOC: Config.Ftp.telnet
5917 The FTP protocol is officially defined to use the telnet protocol
5918 as transport channel for the control connection. However, many
5919 implementations are broken and does not respect this aspect of
5922 If you have trouble accessing files with ASCII code 255 in the
5923 path or similar problems involving this ASCII code you can
5924 try setting this directive to off. If that helps, report to the
5925 operator of the FTP server in question that their FTP server
5926 is broken and does not follow the FTP standard.
5930 OPTIONS FOR EXTERNAL SUPPORT PROGRAMS
5931 -----------------------------------------------------------------------------
5936 DEFAULT: @DEFAULT_DISKD@
5937 LOC: Config.Program.diskd
5939 Specify the location of the diskd executable.
5940 Note this is only useful if you have compiled in
5941 diskd as one of the store io modules.
5944 NAME: unlinkd_program
5947 DEFAULT: @DEFAULT_UNLINKD@
5948 LOC: Config.Program.unlinkd
5950 Specify the location of the executable for file deletion process.
5953 NAME: pinger_program
5956 DEFAULT: @DEFAULT_PINGER@
5959 Specify the location of the executable for the pinger process.
5968 Control whether the pinger is active at run-time.
5969 Enables turning ICMP pinger on and off with a simple
5970 squid -k reconfigure.
5975 OPTIONS FOR URL REWRITING
5976 -----------------------------------------------------------------------------
5979 NAME: url_rewrite_program redirect_program
5981 LOC: Config.Program.redirect
5984 The name and command line parameters of an admin-provided executable
5985 for redirecting clients or adjusting/replacing client request URLs.
5987 This helper is consulted after the received request is cleared by
5988 http_access and adapted using eICAP/ICAP services (if any). If the
5989 helper does not redirect the client, Squid checks adapted_http_access
5990 and may consult the cache or forward the request to the next hop.
5993 For each request, the helper gets one line in the following format:
5995 [channel-ID <SP>] request-URL [<SP> extras] <NL>
5997 Use url_rewrite_extras to configure what Squid sends as 'extras'.
6000 The helper must reply to each query using a single line:
6002 [channel-ID <SP>] result [<SP> kv-pairs] <NL>
6004 The result section must match exactly one of the following outcomes:
6006 OK [status=30N] url="..."
6008 Redirect the client to a URL supplied in the 'url' parameter.
6009 Optional 'status' specifies the status code to send to the
6010 client in Squid's HTTP redirect response. It must be one of
6011 the standard HTTP redirect status codes: 301, 302, 303, 307,
6012 or 308. When no specific status is requested, Squid uses 302.
6014 OK rewrite-url="..."
6016 Replace the current request URL with the one supplied in the
6017 'rewrite-url' parameter. Squid fetches the resource specified
6018 by the new URL and forwards the received response (or its
6019 cached copy) to the client.
6021 WARNING: Avoid rewriting URLs! When possible, redirect the
6022 client using an "OK url=..." helper response instead.
6023 Rewriting URLs may create inconsistent requests and/or break
6024 synchronization between internal client and origin server
6025 states, especially when URLs or other message parts contain
6026 snippets of that state. For example, Squid does not adjust
6027 Location headers and embedded URLs after the helper rewrites
6031 Keep the client request intact.
6034 Keep the client request intact.
6037 A helper problem that should be reported to the Squid admin
6038 via a level-1 cache.log message. The 'message' parameter is
6039 reserved for specifying the log message.
6041 In addition to the kv-pairs mentioned above, Squid also understands
6042 the following optional kv-pairs in URL rewriter responses:
6045 Associates a TAG with the client TCP connection.
6047 The clt_conn_tag=TAG pair is treated as a regular transaction
6048 annotation for the current request and also annotates future
6049 requests on the same client connection. A helper may update
6050 the TAG during subsequent requests by returning a new kv-pair.
6053 Helper messages contain the channel-ID part if and only if the
6054 url_rewrite_children directive specifies positive concurrency. As a
6055 channel-ID value, Squid sends a number between 0 and concurrency-1.
6056 The helper must echo back the received channel-ID in its response.
6058 By default, Squid does not use a URL rewriter.
6061 NAME: url_rewrite_children redirect_children
6062 TYPE: HelperChildConfig
6063 DEFAULT: 20 startup=0 idle=1 concurrency=0
6064 LOC: Config.redirectChildren
6066 Specifies the maximum number of redirector processes that Squid may
6067 spawn (numberofchildren) and several related options. Using too few of
6068 these helper processes (a.k.a. "helpers") creates request queues.
6069 Using too many helpers wastes your system resources.
6071 Usage: numberofchildren [option]...
6073 The startup= and idle= options allow some measure of skew in your
6078 Sets a minimum of how many processes are to be spawned when Squid
6079 starts or reconfigures. When set to zero the first request will
6080 cause spawning of the first child process to handle it.
6082 Starting too few will cause an initial slowdown in traffic as Squid
6083 attempts to simultaneously spawn enough processes to cope.
6087 Sets a minimum of how many processes Squid is to try and keep available
6088 at all times. When traffic begins to rise above what the existing
6089 processes can handle this many more will be spawned up to the maximum
6090 configured. A minimum setting of 1 is required.
6094 The number of requests each redirector helper can handle in
6095 parallel. Defaults to 0 which indicates the redirector
6096 is a old-style single threaded redirector.
6098 When this directive is set to a value >= 1 then the protocol
6099 used to communicate with the helper is modified to include
6100 an ID in front of the request/response. The ID from the request
6101 must be echoed back with the response to that request.
6105 Sets the maximum number of queued requests. A request is queued when
6106 no existing child can accept it due to concurrency limit and no new
6107 child can be started due to numberofchildren limit. The default
6108 maximum is zero if url_rewrite_bypass is enabled and
6109 2*numberofchildren otherwise. If the queued requests exceed queue size
6110 and redirector_bypass configuration option is set, then redirector is
6111 bypassed. Otherwise, Squid is allowed to temporarily exceed the
6112 configured maximum, marking the affected helper as "overloaded". If
6113 the helper overload lasts more than 3 minutes, the action prescribed
6114 by the on-persistent-overload option applies.
6116 on-persistent-overload=action
6118 Specifies Squid reaction to a new helper request arriving when the helper
6119 has been overloaded for more that 3 minutes already. The number of queued
6120 requests determines whether the helper is overloaded (see the queue-size
6123 Two actions are supported:
6125 die Squid worker quits. This is the default behavior.
6127 ERR Squid treats the helper request as if it was
6128 immediately submitted, and the helper immediately
6129 replied with an ERR response. This action has no effect
6130 on the already queued and in-progress helper requests.
6133 NAME: url_rewrite_host_header redirect_rewrites_host_header
6136 LOC: Config.onoff.redir_rewrites_host
6138 To preserve same-origin security policies in browsers and
6139 prevent Host: header forgery by redirectors Squid rewrites
6140 any Host: header in redirected requests.
6142 If you are running an accelerator this may not be a wanted
6143 effect of a redirector. This directive enables you disable
6144 Host: alteration in reverse-proxy traffic.
6146 WARNING: Entries are cached on the result of the URL rewriting
6147 process, so be careful if you have domain-virtual hosts.
6149 WARNING: Squid and other software verifies the URL and Host
6150 are matching, so be careful not to relay through other proxies
6151 or inspecting firewalls with this disabled.
6154 NAME: url_rewrite_access redirector_access
6157 DEFAULT_DOC: Allow, unless rules exist in squid.conf.
6158 LOC: Config.accessList.redirector
6160 If defined, this access list specifies which requests are
6161 sent to the redirector processes.
6163 This clause supports both fast and slow acl types.
6164 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
6167 NAME: url_rewrite_bypass redirector_bypass
6169 LOC: Config.onoff.redirector_bypass
6172 When this is 'on', a request will not go through the
6173 redirector if all the helpers are busy. If this is 'off' and the
6174 redirector queue grows too large, the action is prescribed by the
6175 on-persistent-overload option. You should only enable this if the
6176 redirectors are not critical to your caching system. If you use
6177 redirectors for access control, and you enable this option,
6178 users may have access to pages they should not
6179 be allowed to request.
6181 Enabling this option sets the default url_rewrite_children queue-size
6185 NAME: url_rewrite_extras
6186 TYPE: TokenOrQuotedString
6187 LOC: Config.redirector_extras
6188 DEFAULT: "%>a/%>A %un %>rm myip=%la myport=%lp"
6190 Specifies a string to be append to request line format for the
6191 rewriter helper. "Quoted" format values may contain spaces and
6192 logformat %macros. In theory, any logformat %macro can be used.
6193 In practice, a %macro expands as a dash (-) if the helper request is
6194 sent before the required macro information is available to Squid.
6197 NAME: url_rewrite_timeout
6198 TYPE: UrlHelperTimeout
6199 LOC: Config.onUrlRewriteTimeout
6201 DEFAULT_DOC: Squid waits for the helper response forever
6203 Squid times active requests to redirector. The timeout value and Squid
6204 reaction to a timed out request are configurable using the following
6207 url_rewrite_timeout timeout time-units on_timeout=<action> [response=<quoted-response>]
6209 supported timeout actions:
6210 fail Squid return a ERR_GATEWAY_FAILURE error page
6212 bypass Do not re-write the URL
6214 retry Send the lookup to the helper again
6216 use_configured_response
6217 Use the <quoted-response> as helper response
6221 OPTIONS FOR STORE ID
6222 -----------------------------------------------------------------------------
6225 NAME: store_id_program storeurl_rewrite_program
6227 LOC: Config.Program.store_id
6230 Specify the location of the executable StoreID helper to use.
6231 Since they can perform almost any function there isn't one included.
6233 For each requested URL, the helper will receive one line with the format
6235 [channel-ID <SP>] URL [<SP> extras]<NL>
6238 After processing the request the helper must reply using the following format:
6240 [channel-ID <SP>] result [<SP> kv-pairs]
6242 The result code can be:
6245 Use the StoreID supplied in 'store-id='.
6248 The default is to use HTTP request URL as the store ID.
6251 An internal error occurred in the helper, preventing
6252 a result being identified.
6254 In addition to the above kv-pairs Squid also understands the following
6255 optional kv-pairs received from URL rewriters:
6257 Associates a TAG with the client TCP connection.
6258 Please see url_rewrite_program related documentation for this
6261 Helper programs should be prepared to receive and possibly ignore
6262 additional whitespace-separated tokens on each input line.
6264 When using the concurrency= option the protocol is changed by
6265 introducing a query channel tag in front of the request/response.
6266 The query channel tag is a number between 0 and concurrency-1.
6267 This value must be echoed back unchanged to Squid as the first part
6268 of the response relating to its request.
6270 NOTE: when using StoreID refresh_pattern will apply to the StoreID
6271 returned from the helper and not the URL.
6273 WARNING: Wrong StoreID value returned by a careless helper may result
6274 in the wrong cached response returned to the user.
6276 By default, a StoreID helper is not used.
6279 NAME: store_id_extras
6280 TYPE: TokenOrQuotedString
6281 LOC: Config.storeId_extras
6282 DEFAULT: "%>a/%>A %un %>rm myip=%la myport=%lp"
6284 Specifies a string to be append to request line format for the
6285 StoreId helper. "Quoted" format values may contain spaces and
6286 logformat %macros. In theory, any logformat %macro can be used.
6287 In practice, a %macro expands as a dash (-) if the helper request is
6288 sent before the required macro information is available to Squid.
6291 NAME: store_id_children storeurl_rewrite_children
6292 TYPE: HelperChildConfig
6293 DEFAULT: 20 startup=0 idle=1 concurrency=0
6294 LOC: Config.storeIdChildren
6296 Specifies the maximum number of StoreID helper processes that Squid
6297 may spawn (numberofchildren) and several related options. Using
6298 too few of these helper processes (a.k.a. "helpers") creates request
6299 queues. Using too many helpers wastes your system resources.
6301 Usage: numberofchildren [option]...
6303 The startup= and idle= options allow some measure of skew in your
6308 Sets a minimum of how many processes are to be spawned when Squid
6309 starts or reconfigures. When set to zero the first request will
6310 cause spawning of the first child process to handle it.
6312 Starting too few will cause an initial slowdown in traffic as Squid
6313 attempts to simultaneously spawn enough processes to cope.
6317 Sets a minimum of how many processes Squid is to try and keep available
6318 at all times. When traffic begins to rise above what the existing
6319 processes can handle this many more will be spawned up to the maximum
6320 configured. A minimum setting of 1 is required.
6324 The number of requests each storeID helper can handle in
6325 parallel. Defaults to 0 which indicates the helper
6326 is a old-style single threaded program.
6328 When this directive is set to a value >= 1 then the protocol
6329 used to communicate with the helper is modified to include
6330 an ID in front of the request/response. The ID from the request
6331 must be echoed back with the response to that request.
6335 Sets the maximum number of queued requests to N. A request is queued
6336 when no existing child can accept it due to concurrency limit and no
6337 new child can be started due to numberofchildren limit. The default
6338 maximum is 2*numberofchildren. If the queued requests exceed queue
6339 size and redirector_bypass configuration option is set, then
6340 redirector is bypassed. Otherwise, Squid is allowed to temporarily
6341 exceed the configured maximum, marking the affected helper as
6342 "overloaded". If the helper overload lasts more than 3 minutes, the
6343 action prescribed by the on-persistent-overload option applies.
6345 on-persistent-overload=action
6347 Specifies Squid reaction to a new helper request arriving when the helper
6348 has been overloaded for more that 3 minutes already. The number of queued
6349 requests determines whether the helper is overloaded (see the queue-size
6352 Two actions are supported:
6354 die Squid worker quits. This is the default behavior.
6356 ERR Squid treats the helper request as if it was
6357 immediately submitted, and the helper immediately
6358 replied with an ERR response. This action has no effect
6359 on the already queued and in-progress helper requests.
6362 NAME: store_id_access storeurl_rewrite_access
6365 DEFAULT_DOC: Allow, unless rules exist in squid.conf.
6366 LOC: Config.accessList.store_id
6368 If defined, this access list specifies which requests are
6369 sent to the StoreID processes. By default all requests
6372 This clause supports both fast and slow acl types.
6373 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
6376 NAME: store_id_bypass storeurl_rewrite_bypass
6378 LOC: Config.onoff.store_id_bypass
6381 When this is 'on', a request will not go through the
6382 helper if all helpers are busy. If this is 'off' and the helper
6383 queue grows too large, the action is prescribed by the
6384 on-persistent-overload option. You should only enable this if the
6385 helpers are not critical to your caching system. If you use
6386 helpers for critical caching components, and you enable this
6387 option, users may not get objects from cache.
6388 This options sets default queue-size option of the store_id_children
6393 OPTIONS FOR TUNING THE CACHE
6394 -----------------------------------------------------------------------------
6397 NAME: cache no_cache
6400 DEFAULT_DOC: By default, this directive is unused and has no effect.
6401 LOC: Config.accessList.noCache
6403 Requests denied by this directive will not be served from the cache
6404 and their responses will not be stored in the cache. This directive
6405 has no effect on other transactions and on already cached responses.
6407 This clause supports both fast and slow acl types.
6408 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
6410 This and the two other similar caching directives listed below are
6411 checked at different transaction processing stages, have different
6412 access to response information, affect different cache operations,
6413 and differ in slow ACLs support:
6415 * cache: Checked before Squid makes a hit/miss determination.
6416 No access to reply information!
6417 Denies both serving a hit and storing a miss.
6418 Supports both fast and slow ACLs.
6419 * send_hit: Checked after a hit was detected.
6420 Has access to reply (hit) information.
6421 Denies serving a hit only.
6422 Supports fast ACLs only.
6423 * store_miss: Checked before storing a cachable miss.
6424 Has access to reply (miss) information.
6425 Denies storing a miss only.
6426 Supports fast ACLs only.
6428 If you are not sure which of the three directives to use, apply the
6429 following decision logic:
6431 * If your ACL(s) are of slow type _and_ need response info, redesign.
6432 Squid does not support that particular combination at this time.
6434 * If your directive ACL(s) are of slow type, use "cache"; and/or
6435 * if your directive ACL(s) need no response info, use "cache".
6437 * If you do not want the response cached, use store_miss; and/or
6438 * if you do not want a hit on a cached response, use send_hit.
6444 DEFAULT_DOC: By default, this directive is unused and has no effect.
6445 LOC: Config.accessList.sendHit
6447 Responses denied by this directive will not be served from the cache
6448 (but may still be cached, see store_miss). This directive has no
6449 effect on the responses it allows and on the cached objects. This
6450 directive is applied to both regular from-cache responses and responses
6451 reused by collapsed requests (see collapsed_forwarding).
6453 Please see the "cache" directive for a summary of differences among
6454 store_miss, send_hit, and cache directives.
6456 Unlike the "cache" directive, send_hit only supports fast acl
6457 types. See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
6461 # apply custom Store ID mapping to some URLs
6462 acl MapMe dstdomain .c.example.com
6463 store_id_program ...
6464 store_id_access allow MapMe
6466 # but prevent caching of special responses
6467 # such as 302 redirects that cause StoreID loops
6468 acl Ordinary http_status 200-299
6469 store_miss deny MapMe !Ordinary
6471 # and do not serve any previously stored special responses
6472 # from the cache (in case they were already cached before
6473 # the above store_miss rule was in effect).
6474 send_hit deny MapMe !Ordinary
6480 DEFAULT_DOC: By default, this directive is unused and has no effect.
6481 LOC: Config.accessList.storeMiss
6483 Responses denied by this directive will not be cached (but may still
6484 be served from the cache, see send_hit). This directive has no
6485 effect on the responses it allows and on the already cached responses.
6487 Please see the "cache" directive for a summary of differences among
6488 store_miss, send_hit, and cache directives. See the
6489 send_hit directive for a usage example.
6491 Unlike the "cache" directive, store_miss only supports fast acl
6492 types. See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
6498 LOC: Config.maxStale
6501 This option puts an upper limit on how stale content Squid
6502 will serve from the cache if cache validation fails.
6503 Can be overridden by the refresh_pattern max-stale option.
6506 NAME: refresh_pattern
6507 TYPE: refreshpattern
6511 usage: refresh_pattern [-i] regex min percent max [options]
6513 By default, regular expressions are CASE-SENSITIVE. To make
6514 them case-insensitive, use the -i option.
6516 'Min' is the time (in minutes) an object without an explicit
6517 expiry time should be considered fresh. The recommended
6518 value is 0, any higher values may cause dynamic applications
6519 to be erroneously cached unless the application designer
6520 has taken the appropriate actions.
6522 'Percent' is used to compute the max-age value for responses
6523 with a Last-Modified header and no Cache-Control:max-age nor Expires.
6524 Cache-Control:max-age = ( Date - Last-Modified ) * percent
6526 'Max' is an upper limit on how long objects without an explicit
6527 expiry time will be considered fresh. The value is also used
6528 to form Cache-Control: max-age header for a request sent from
6529 Squid to origin/parent.
6531 options: override-expire
6541 override-expire enforces min age even if the server
6542 sent an explicit expiry time (e.g., with the
6543 Expires: header or Cache-Control: max-age). Doing this
6544 VIOLATES the HTTP standard. Enabling this feature
6545 could make you liable for problems which it causes.
6547 Note: override-expire does not enforce staleness - it only extends
6548 freshness / min. If the server returns a Expires time which
6549 is longer than your max time, Squid will still consider
6550 the object fresh for that period of time.
6552 override-lastmod enforces min age even on objects
6553 that were modified recently.
6555 reload-into-ims changes a client no-cache or ``reload''
6556 request for a cached entry into a conditional request using
6557 If-Modified-Since and/or If-None-Match headers, provided the
6558 cached entry has a Last-Modified and/or a strong ETag header.
6559 Doing this VIOLATES the HTTP standard. Enabling this feature
6560 could make you liable for problems which it causes.
6562 ignore-reload ignores a client no-cache or ``reload''
6563 header. Doing this VIOLATES the HTTP standard. Enabling
6564 this feature could make you liable for problems which
6567 ignore-no-store ignores any ``Cache-control: no-store''
6568 headers received from a server. Doing this VIOLATES
6569 the HTTP standard. Enabling this feature could make you
6570 liable for problems which it causes.
6572 ignore-private ignores any ``Cache-control: private''
6573 headers received from a server. Doing this VIOLATES
6574 the HTTP standard. Enabling this feature could make you
6575 liable for problems which it causes.
6577 refresh-ims causes squid to contact the origin server
6578 when a client issues an If-Modified-Since request. This
6579 ensures that the client will receive an updated version
6580 if one is available.
6582 store-stale stores responses even if they don't have explicit
6583 freshness or a validator (i.e., Last-Modified or an ETag)
6584 present, or if they're already stale. By default, Squid will
6585 not cache such responses because they usually can't be
6586 reused. Note that such responses will be stale by default.
6588 max-stale=NN provide a maximum staleness factor. Squid won't
6589 serve objects more stale than this even if it failed to
6590 validate the object. Default: use the max_stale global limit.
6592 Basically a cached object is:
6594 FRESH if expire > now, else STALE
6596 FRESH if lm-factor < percent, else STALE
6600 The refresh_pattern lines are checked in the order listed here.
6601 The first entry which matches is used. If none of the entries
6602 match the default will be used.
6604 Note, you must uncomment all the default lines if you want
6605 to change one. The default setting is only active if none is
6611 # Add any of your own refresh_pattern entries above these.
6613 refresh_pattern ^ftp: 1440 20% 10080
6614 refresh_pattern -i (/cgi-bin/|\?) 0 0% 0
6615 refresh_pattern . 0 20% 4320
6619 NAME: quick_abort_min
6623 LOC: Config.quickAbort.min
6626 NAME: quick_abort_max
6630 LOC: Config.quickAbort.max
6633 NAME: quick_abort_pct
6637 LOC: Config.quickAbort.pct
6639 Continuing to download a cachable response after its request is aborted is
6640 going to waste resources if the received response is not requested again.
6641 On the other hand, aborting an in-progress download may effectively waste
6642 (already spent) resources if the received cachable response is requested
6643 again. Such waste is especially noticeable when, for example, impatient
6644 users repeatedly request and then abort slow downloads. To balance these
6645 trade-offs when a request is aborted during response download, Squid may
6646 check quick_abort_* directives to decide whether to finish the retrieval:
6648 If the transfer has less than 'quick_abort_min' KB remaining,
6649 it will finish the retrieval.
6651 If the transfer has more than 'quick_abort_max' KB remaining,
6652 it will abort the retrieval.
6654 If more than 'quick_abort_pct' of the transfer has completed,
6655 it will finish the retrieval.
6657 If you do not want any retrieval to continue after the client
6658 has aborted, set both 'quick_abort_min' and 'quick_abort_max'
6661 If you want retrievals to always continue if they are being
6662 cached set 'quick_abort_min' to '-1 KB'.
6664 Many other conditions affect Squid decision to abort or continue download.
6665 For example, Squid continues to download responses that feed other
6666 requests but aborts responses with unknown body length.
6669 NAME: read_ahead_gap
6670 COMMENT: buffer-size
6672 LOC: Config.readAheadGap
6675 The amount of data the cache will buffer ahead of what has been
6676 sent to the client when retrieving an object from another server.
6680 IFDEF: USE_HTTP_VIOLATIONS
6683 LOC: Config.negativeTtl
6686 Set the Default Time-to-Live (TTL) for failed requests.
6687 Certain types of failures (such as "connection refused" and
6688 "404 Not Found") are able to be negatively-cached for a short time.
6689 Modern web servers should provide Expires: header, however if they
6690 do not this can provide a minimum TTL.
6691 The default is not to cache errors with unknown expiry details.
6693 Note that this is different from negative caching of DNS lookups.
6695 WARNING: Doing this VIOLATES the HTTP standard. Enabling
6696 this feature could make you liable for problems which it
6700 NAME: positive_dns_ttl
6703 LOC: Config.positiveDnsTtl
6706 Upper limit on how long Squid will cache positive DNS responses.
6707 Default is 6 hours (360 minutes). This directive must be set
6708 larger than negative_dns_ttl.
6711 NAME: negative_dns_ttl
6714 LOC: Config.negativeDnsTtl
6717 Time-to-Live (TTL) for negative caching of failed DNS lookups.
6718 This also sets the lower cache limit on positive lookups.
6719 Minimum value is 1 second, and it is not recommendable to go
6720 much below 10 seconds.
6723 NAME: range_offset_limit
6724 COMMENT: size [acl acl...]
6726 LOC: Config.rangeOffsetLimit
6729 usage: (size) [units] [[!]aclname]
6731 Sets an upper limit on how far (number of bytes) into the file
6732 a Range request may be to cause Squid to prefetch the whole file.
6733 If beyond this limit, Squid forwards the Range request as it is and
6734 the result is NOT cached.
6736 This is to stop a far ahead range request (lets say start at 17MB)
6737 from making Squid fetch the whole object up to that point before
6738 sending anything to the client.
6740 Multiple range_offset_limit lines may be specified, and they will
6741 be searched from top to bottom on each request until a match is found.
6742 The first match found will be used. If no line matches a request, the
6743 default limit of 0 bytes will be used.
6745 'size' is the limit specified as a number of units.
6747 'units' specifies whether to use bytes, KB, MB, etc.
6748 If no units are specified bytes are assumed.
6750 A size of 0 causes Squid to never fetch more than the
6751 client requested. (default)
6753 A size of 'none' causes Squid to always fetch the object from the
6754 beginning so it may cache the result. (2.0 style)
6756 'aclname' is the name of a defined ACL.
6758 NP: Using 'none' as the byte value here will override any quick_abort settings
6759 that may otherwise apply to the range request. The range request will
6760 be fully fetched from start to finish regardless of the client
6761 actions. This affects bandwidth usage.
6764 NAME: minimum_expiry_time
6767 LOC: Config.minimum_expiry_time
6770 The minimum caching time according to (Expires - Date)
6771 headers Squid honors if the object can't be revalidated.
6772 The default is 60 seconds.
6774 In reverse proxy environments it might be desirable to honor
6775 shorter object lifetimes. It is most likely better to make
6776 your server return a meaningful Last-Modified header however.
6779 NAME: store_avg_object_size
6783 LOC: Config.Store.avgObjectSize
6785 Average object size, used to estimate number of objects your
6786 cache can hold. The default is 13 KB.
6788 This is used to pre-seed the cache index memory allocation to
6789 reduce expensive reallocate operations while handling clients
6790 traffic. Too-large values may result in memory allocation during
6791 peak traffic, too-small values will result in wasted memory.
6793 Check the cache manager 'info' report metrics for the real
6794 object sizes seen by your Squid before tuning this.
6797 NAME: store_objects_per_bucket
6800 LOC: Config.Store.objectsPerBucket
6802 Target number of objects per bucket in the store hash table.
6803 Lowering this value increases the total number of buckets and
6804 also the storage maintenance rate. The default is 20.
6809 -----------------------------------------------------------------------------
6812 NAME: request_header_max_size
6816 LOC: Config.maxRequestHeaderSize
6818 This directives limits the header size of a received HTTP request
6819 (including request-line). Increasing this limit beyond its 64 KB default
6820 exposes certain old Squid code to various denial-of-service attacks. This
6821 limit also applies to received FTP commands.
6823 This limit has no direct affect on Squid memory consumption.
6825 Squid does not check this limit when sending requests.
6828 NAME: reply_header_max_size
6832 LOC: Config.maxReplyHeaderSize
6834 This directives limits the header size of a received HTTP response
6835 (including status-line). Increasing this limit beyond its 64 KB default
6836 exposes certain old Squid code to various denial-of-service attacks. This
6837 limit also applies to FTP command responses.
6839 Squid also checks this limit when loading hit responses from disk cache.
6841 Squid does not check this limit when sending responses.
6844 NAME: request_body_max_size
6848 DEFAULT_DOC: No limit.
6849 LOC: Config.maxRequestBodySize
6851 This specifies the maximum size for an HTTP request body.
6852 In other words, the maximum size of a PUT/POST request.
6853 A user who attempts to send a request with a body larger
6854 than this limit receives an "Invalid Request" error message.
6855 If you set this parameter to a zero (the default), there will
6856 be no limit imposed.
6858 See also client_request_buffer_max_size for an alternative
6859 limitation on client uploads which can be configured.
6862 NAME: client_request_buffer_max_size
6866 LOC: Config.maxRequestBufferSize
6868 This specifies the maximum buffer size of a client request.
6869 It prevents squid eating too much memory when somebody uploads
6874 IFDEF: USE_HTTP_VIOLATIONS
6877 DEFAULT_DOC: Obey RFC 2616.
6878 LOC: Config.accessList.brokenPosts
6880 A list of ACL elements which, if matched, causes Squid to send
6881 an extra CRLF pair after the body of a PUT/POST request.
6883 Some HTTP servers has broken implementations of PUT/POST,
6884 and rely on an extra CRLF pair sent by some WWW clients.
6886 Quote from RFC2616 section 4.1 on this matter:
6888 Note: certain buggy HTTP/1.0 client implementations generate an
6889 extra CRLF's after a POST request. To restate what is explicitly
6890 forbidden by the BNF, an HTTP/1.1 client must not preface or follow
6891 a request with an extra CRLF.
6893 This clause only supports fast acl types.
6894 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
6897 acl buggy_server url_regex ^http://....
6898 broken_posts allow buggy_server
6901 NAME: adaptation_uses_indirect_client icap_uses_indirect_client
6904 IFDEF: FOLLOW_X_FORWARDED_FOR&&USE_ADAPTATION
6906 LOC: Adaptation::Config::use_indirect_client
6908 Controls whether the indirect client IP address (instead of the direct
6909 client IP address) is passed to adaptation services.
6911 See also: follow_x_forwarded_for adaptation_send_client_ip
6915 IFDEF: USE_HTTP_VIOLATIONS
6919 LOC: Config.onoff.via
6921 If set (default), Squid will include a Via header in requests and
6922 replies as required by RFC2616.
6925 NAME: vary_ignore_expire
6928 LOC: Config.onoff.vary_ignore_expire
6931 Many HTTP servers supporting Vary gives such objects
6932 immediate expiry time with no cache-control header
6933 when requested by a HTTP/1.0 client. This option
6934 enables Squid to ignore such expiry times until
6935 HTTP/1.1 is fully implemented.
6937 WARNING: If turned on this may eventually cause some
6938 varying objects not intended for caching to get cached.
6941 NAME: request_header_access
6942 IFDEF: USE_HTTP_VIOLATIONS
6943 TYPE: http_header_access
6944 LOC: Config.request_header_access
6946 DEFAULT_DOC: No limits.
6948 Usage: request_header_access header_name allow|deny [!]aclname ...
6950 WARNING: Doing this VIOLATES the HTTP standard. Enabling
6951 this feature could make you liable for problems which it
6954 This option replaces the old 'anonymize_headers' and the
6955 older 'http_anonymizer' option with something that is much
6956 more configurable. A list of ACLs for each header name allows
6957 removal of specific header fields under specific conditions.
6959 This option only applies to outgoing HTTP request headers (i.e.,
6960 headers sent by Squid to the next HTTP hop such as a cache peer
6961 or an origin server). The option has no effect during cache hit
6962 detection. The equivalent adaptation vectoring point in ICAP
6963 terminology is post-cache REQMOD.
6965 The option is applied to individual outgoing request header
6966 fields. For each request header field F, Squid uses the first
6967 qualifying sets of request_header_access rules:
6969 1. Rules with header_name equal to F's name.
6970 2. Rules with header_name 'Other', provided F's name is not
6971 on the hard-coded list of commonly used HTTP header names.
6972 3. Rules with header_name 'All'.
6974 Within that qualifying rule set, rule ACLs are checked as usual.
6975 If ACLs of an "allow" rule match, the header field is allowed to
6976 go through as is. If ACLs of a "deny" rule match, the header is
6977 removed and request_header_replace is then checked to identify
6978 if the removed header has a replacement. If no rules within the
6979 set have matching ACLs, the header field is left as is.
6981 For example, to achieve the same behavior as the old
6982 'http_anonymizer standard' option, you should use:
6984 request_header_access From deny all
6985 request_header_access Referer deny all
6986 request_header_access User-Agent deny all
6988 Or, to reproduce the old 'http_anonymizer paranoid' feature
6991 request_header_access Authorization allow all
6992 request_header_access Proxy-Authorization allow all
6993 request_header_access Cache-Control allow all
6994 request_header_access Content-Length allow all
6995 request_header_access Content-Type allow all
6996 request_header_access Date allow all
6997 request_header_access Host allow all
6998 request_header_access If-Modified-Since allow all
6999 request_header_access Pragma allow all
7000 request_header_access Accept allow all
7001 request_header_access Accept-Charset allow all
7002 request_header_access Accept-Encoding allow all
7003 request_header_access Accept-Language allow all
7004 request_header_access Connection allow all
7005 request_header_access All deny all
7007 HTTP reply headers are controlled with the reply_header_access directive.
7009 By default, all headers are allowed (no anonymizing is performed).
7012 NAME: reply_header_access
7013 IFDEF: USE_HTTP_VIOLATIONS
7014 TYPE: http_header_access
7015 LOC: Config.reply_header_access
7017 DEFAULT_DOC: No limits.
7019 Usage: reply_header_access header_name allow|deny [!]aclname ...
7021 WARNING: Doing this VIOLATES the HTTP standard. Enabling
7022 this feature could make you liable for problems which it
7025 This option only applies to reply headers, i.e., from the
7026 server to the client.
7028 This is the same as request_header_access, but in the other
7029 direction. Please see request_header_access for detailed
7032 For example, to achieve the same behavior as the old
7033 'http_anonymizer standard' option, you should use:
7035 reply_header_access Server deny all
7036 reply_header_access WWW-Authenticate deny all
7037 reply_header_access Link deny all
7039 Or, to reproduce the old 'http_anonymizer paranoid' feature
7042 reply_header_access Allow allow all
7043 reply_header_access WWW-Authenticate allow all
7044 reply_header_access Proxy-Authenticate allow all
7045 reply_header_access Cache-Control allow all
7046 reply_header_access Content-Encoding allow all
7047 reply_header_access Content-Length allow all
7048 reply_header_access Content-Type allow all
7049 reply_header_access Date allow all
7050 reply_header_access Expires allow all
7051 reply_header_access Last-Modified allow all
7052 reply_header_access Location allow all
7053 reply_header_access Pragma allow all
7054 reply_header_access Content-Language allow all
7055 reply_header_access Retry-After allow all
7056 reply_header_access Title allow all
7057 reply_header_access Content-Disposition allow all
7058 reply_header_access Connection allow all
7059 reply_header_access All deny all
7061 HTTP request headers are controlled with the request_header_access directive.
7063 By default, all headers are allowed (no anonymizing is
7067 NAME: request_header_replace header_replace
7068 IFDEF: USE_HTTP_VIOLATIONS
7069 TYPE: http_header_replace
7070 LOC: Config.request_header_access
7073 Usage: request_header_replace header_name message
7074 Example: request_header_replace User-Agent Nutscrape/1.0 (CP/M; 8-bit)
7076 This option allows you to change the contents of headers
7077 denied with request_header_access above, by replacing them
7078 with some fixed string.
7080 This only applies to request headers, not reply headers.
7082 By default, headers are removed if denied.
7085 NAME: reply_header_replace
7086 IFDEF: USE_HTTP_VIOLATIONS
7087 TYPE: http_header_replace
7088 LOC: Config.reply_header_access
7091 Usage: reply_header_replace header_name message
7092 Example: reply_header_replace Server Foo/1.0
7094 This option allows you to change the contents of headers
7095 denied with reply_header_access above, by replacing them
7096 with some fixed string.
7098 This only applies to reply headers, not request headers.
7100 By default, headers are removed if denied.
7103 NAME: request_header_add
7104 TYPE: HeaderWithAclList
7105 LOC: Config.request_header_add
7108 Usage: request_header_add field-name field-value [ acl ... ]
7109 Example: request_header_add X-Client-CA "CA=%ssl::>cert_issuer" all
7111 This option adds header fields to outgoing HTTP requests (i.e.,
7112 request headers sent by Squid to the next HTTP hop such as a
7113 cache peer or an origin server). The option has no effect during
7114 cache hit detection. The equivalent adaptation vectoring point
7115 in ICAP terminology is post-cache REQMOD.
7117 Field-name is a token specifying an HTTP header name. If a
7118 standard HTTP header name is used, Squid does not check whether
7119 the new header conflicts with any existing headers or violates
7120 HTTP rules. If the request to be modified already contains a
7121 field with the same name, the old field is preserved but the
7122 header field values are not merged.
7124 Field-value is either a token or a quoted string. If quoted
7125 string format is used, then the surrounding quotes are removed
7126 while escape sequences and %macros are processed.
7128 One or more Squid ACLs may be specified to restrict header
7129 injection to matching requests. As always in squid.conf, all
7130 ACLs in the ACL list must be satisfied for the insertion to
7131 happen. The request_header_add supports fast ACLs only.
7133 See also: reply_header_add.
7136 NAME: reply_header_add
7137 TYPE: HeaderWithAclList
7138 LOC: Config.reply_header_add
7141 Usage: reply_header_add field-name field-value [ acl ... ]
7142 Example: reply_header_add X-Client-CA "CA=%ssl::>cert_issuer" all
7144 This option adds header fields to outgoing HTTP responses (i.e., response
7145 headers delivered by Squid to the client). This option has no effect on
7146 cache hit detection. The equivalent adaptation vectoring point in
7147 ICAP terminology is post-cache RESPMOD. This option does not apply to
7148 successful CONNECT replies.
7150 Field-name is a token specifying an HTTP header name. If a
7151 standard HTTP header name is used, Squid does not check whether
7152 the new header conflicts with any existing headers or violates
7153 HTTP rules. If the response to be modified already contains a
7154 field with the same name, the old field is preserved but the
7155 header field values are not merged.
7157 Field-value is either a token or a quoted string. If quoted
7158 string format is used, then the surrounding quotes are removed
7159 while escape sequences and %macros are processed.
7161 One or more Squid ACLs may be specified to restrict header
7162 injection to matching responses. As always in squid.conf, all
7163 ACLs in the ACL list must be satisfied for the insertion to
7164 happen. The reply_header_add option supports fast ACLs only.
7166 See also: request_header_add.
7174 This option used to log custom information about the master
7175 transaction. For example, an admin may configure Squid to log
7176 which "user group" the transaction belongs to, where "user group"
7177 will be determined based on a set of ACLs and not [just]
7178 authentication information.
7179 Values of key/value pairs can be logged using %{key}note macros:
7181 note key value acl ...
7182 logformat myFormat ... %{key}note ...
7184 This clause only supports fast acl types.
7185 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
7188 NAME: relaxed_header_parser
7189 COMMENT: on|off|warn
7191 LOC: Config.onoff.relaxed_header_parser
7194 In the default "on" setting Squid accepts certain forms
7195 of non-compliant HTTP messages where it is unambiguous
7196 what the sending application intended even if the message
7197 is not correctly formatted. The messages is then normalized
7198 to the correct form when forwarded by Squid.
7200 If set to "warn" then a warning will be emitted in cache.log
7201 each time such HTTP error is encountered.
7203 If set to "off" then such HTTP errors will cause the request
7204 or response to be rejected.
7207 NAME: collapsed_forwarding
7210 LOC: Config.onoff.collapsed_forwarding
7213 This option controls whether Squid is allowed to merge multiple
7214 potentially cachable requests for the same URI before Squid knows
7215 whether the response is going to be cachable.
7217 When enabled, instead of forwarding each concurrent request for
7218 the same URL, Squid just sends the first of them. The other, so
7219 called "collapsed" requests, wait for the response to the first
7220 request and, if it happens to be cachable, use that response.
7221 Here, "concurrent requests" means "received after the first
7222 request headers were parsed and before the corresponding response
7223 headers were parsed".
7225 This feature is disabled by default: enabling collapsed
7226 forwarding needlessly delays forwarding requests that look
7227 cachable (when they are collapsed) but then need to be forwarded
7228 individually anyway because they end up being for uncachable
7229 content. However, in some cases, such as acceleration of highly
7230 cachable content with periodic or grouped expiration times, the
7231 gains from collapsing [large volumes of simultaneous refresh
7232 requests] outweigh losses from such delays.
7234 Squid collapses two kinds of requests: regular client requests
7235 received on one of the listening ports and internal "cache
7236 revalidation" requests which are triggered by those regular
7237 requests hitting a stale cached object. Revalidation collapsing
7238 is currently disabled for Squid instances containing SMP-aware
7239 disk or memory caches and for Vary-controlled cached objects.
7241 A response reused by the collapsed request is deemed fresh in that
7242 request processing context -- Squid does not apply refresh_pattern and
7243 internal freshness validation checks to collapsed transactions. Squid
7244 does apply send_hit rules.
7247 NAME: collapsed_forwarding_access
7250 DEFAULT_DOC: Requests may be collapsed if collapsed_forwarding is on.
7251 LOC: Config.accessList.collapsedForwardingAccess
7253 Use this directive to restrict collapsed forwarding to a subset of
7254 eligible requests. The directive is checked for regular HTTP
7255 requests, internal revalidation requests, and HTCP/ICP requests.
7257 collapsed_forwarding_access allow|deny [!]aclname ...
7259 This directive cannot force collapsing. It has no effect on
7260 collapsing unless collapsed_forwarding is 'on', and all other
7261 collapsing preconditions are satisfied.
7263 * A denied request will not collapse, and future transactions will
7264 not collapse on it (even if they are allowed to collapse).
7266 * An allowed request may collapse, or future transactions may
7267 collapse on it (provided they are allowed to collapse).
7269 This directive is evaluated before receiving HTTP response headers
7270 and without access to Squid-to-peer connection (if any).
7272 Only fast ACLs are supported.
7274 See also: collapsed_forwarding.
7277 NAME: shared_transient_entries_limit collapsed_forwarding_shared_entries_limit
7278 COMMENT: (number of entries)
7280 LOC: Config.shared_transient_entries_limit
7283 This directive limits the size of a table used for sharing current
7284 transaction information among SMP workers. A table entry stores meta
7285 information about a single cache entry being delivered to Squid
7286 client(s) by one or more SMP workers. A single table entry consumes
7287 less than 128 shared memory bytes.
7289 The limit should be significantly larger than the number of
7290 concurrent non-collapsed cachable responses leaving Squid. For a
7291 cache that handles less than 5000 concurrent requests, the default
7292 setting of 16384 should be plenty.
7294 Using excessively large values wastes shared memory. Limiting the
7295 table size too much results in hash collisions, leading to lower hit
7296 ratio and missed SMP request collapsing opportunities: Transactions
7297 left without a table entry cannot cache their responses and are
7298 invisible to other concurrent requests for the same resource.
7300 A zero limit is allowed but unsupported. A positive small limit
7301 lowers hit ratio, but zero limit disables a lot of essential
7302 synchronization among SMP workers, leading to HTTP violations (e.g.,
7303 stale hit responses). It also disables shared collapsed forwarding:
7304 A worker becomes unable to collapse its requests on transactions in
7305 other workers, resulting in more trips to the origin server and more
7311 -----------------------------------------------------------------------------
7314 NAME: forward_timeout
7317 LOC: Config.Timeout.forward
7320 This parameter specifies how long Squid should at most attempt in
7321 finding a forwarding path for the request before giving up.
7324 NAME: connect_timeout
7327 LOC: Config.Timeout.connect
7330 This parameter specifies how long to wait for the TCP connect to
7331 the requested server or peer to complete before Squid should
7332 attempt to find another path where to forward the request.
7335 NAME: peer_connect_timeout
7338 LOC: Config.Timeout.peer_connect
7341 This parameter specifies how long to wait for a pending TCP
7342 connection to a peer cache. The default is 30 seconds. You
7343 may also set different timeout values for individual neighbors
7344 with the 'connect-timeout' option on a 'cache_peer' line.
7350 LOC: Config.Timeout.read
7353 Applied on peer server connections.
7355 After each successful read(), the timeout will be extended by this
7356 amount. If no data is read again after this amount of time,
7357 the request is aborted and logged with ERR_READ_TIMEOUT.
7359 The default is 15 minutes.
7365 LOC: Config.Timeout.write
7368 This timeout is tracked for all connections that have data
7369 available for writing and are waiting for the socket to become
7370 ready. After each successful write, the timeout is extended by
7371 the configured amount. If Squid has data to write but the
7372 connection is not ready for the configured duration, the
7373 transaction associated with the connection is terminated. The
7374 default is 15 minutes.
7377 NAME: request_timeout
7379 LOC: Config.Timeout.request
7382 How long to wait for complete HTTP request headers after initial
7383 connection establishment.
7386 NAME: request_start_timeout
7388 LOC: Config.Timeout.request_start_timeout
7391 How long to wait for the first request byte after initial
7392 connection establishment.
7395 NAME: client_idle_pconn_timeout persistent_request_timeout
7397 LOC: Config.Timeout.clientIdlePconn
7400 How long to wait for the next HTTP request on a persistent
7401 client connection after the previous request completes.
7404 NAME: ftp_client_idle_timeout
7406 LOC: Config.Timeout.ftpClientIdle
7409 How long to wait for an FTP request on a connection to Squid ftp_port.
7410 Many FTP clients do not deal with idle connection closures well,
7411 necessitating a longer default timeout than client_idle_pconn_timeout
7412 used for incoming HTTP requests.
7415 NAME: client_lifetime
7418 LOC: Config.Timeout.lifetime
7421 The maximum amount of time a client (browser) is allowed to
7422 remain connected to the cache process. This protects the Cache
7423 from having a lot of sockets (and hence file descriptors) tied up
7424 in a CLOSE_WAIT state from remote clients that go away without
7425 properly shutting down (either because of a network failure or
7426 because of a poor client implementation). The default is one
7429 NOTE: The default value is intended to be much larger than any
7430 client would ever need to be connected to your cache. You
7431 should probably change client_lifetime only as a last resort.
7432 If you seem to have many client connections tying up
7433 filedescriptors, we recommend first tuning the read_timeout,
7434 request_timeout, persistent_request_timeout and quick_abort values.
7437 NAME: pconn_lifetime
7440 LOC: Config.Timeout.pconnLifetime
7443 Desired maximum lifetime of a persistent connection.
7444 When set, Squid will close a now-idle persistent connection that
7445 exceeded configured lifetime instead of moving the connection into
7446 the idle connection pool (or equivalent). No effect on ongoing/active
7447 transactions. Connection lifetime is the time period from the
7448 connection acceptance or opening time until "now".
7450 This limit is useful in environments with long-lived connections
7451 where Squid configuration or environmental factors change during a
7452 single connection lifetime. If unrestricted, some connections may
7453 last for hours and even days, ignoring those changes that should
7454 have affected their behavior or their existence.
7456 Currently, a new lifetime value supplied via Squid reconfiguration
7457 has no effect on already idle connections unless they become busy.
7459 When set to '0' this limit is not used.
7462 NAME: half_closed_clients
7464 LOC: Config.onoff.half_closed_clients
7467 Some clients may shutdown the sending side of their TCP
7468 connections, while leaving their receiving sides open. Sometimes,
7469 Squid can not tell the difference between a half-closed and a
7470 fully-closed TCP connection.
7472 By default, Squid will immediately close client connections when
7473 read(2) returns "no more data to read."
7475 Change this option to 'on' and Squid will keep open connections
7476 until a read(2) or write(2) on the socket returns an error.
7477 This may show some benefits for reverse proxies. But if not
7478 it is recommended to leave OFF.
7481 NAME: server_idle_pconn_timeout pconn_timeout
7483 LOC: Config.Timeout.serverIdlePconn
7486 Timeout for idle persistent connections to servers and other
7490 NAME: shutdown_lifetime
7493 LOC: Config.shutdownLifetime
7496 When SIGTERM or SIGHUP is received, the cache is put into
7497 "shutdown pending" mode until all active sockets are closed.
7498 This value is the lifetime to set for all open descriptors
7499 during shutdown mode. Any active clients after this many
7500 seconds will receive a 'timeout' message.
7504 ADMINISTRATIVE PARAMETERS
7505 -----------------------------------------------------------------------------
7511 LOC: Config.adminEmail
7513 Email-address of local cache manager who will receive
7514 mail if the cache dies. The default is "webmaster".
7520 LOC: Config.EmailFrom
7522 From: email-address for mail sent when the cache dies.
7523 The default is to use 'squid@unique_hostname'.
7525 See also: unique_hostname directive.
7531 LOC: Config.EmailProgram
7533 Email program used to send mail if the cache dies.
7534 The default is "mail". The specified program must comply
7535 with the standard Unix mail syntax:
7536 mail-program recipient < mailfile
7538 Optional command line options can be specified.
7541 NAME: cache_effective_user
7543 DEFAULT: @DEFAULT_CACHE_EFFECTIVE_USER@
7544 LOC: Config.effectiveUser
7546 If you start Squid as root, it will change its effective/real
7547 UID/GID to the user specified below. The default is to change
7548 to UID of @DEFAULT_CACHE_EFFECTIVE_USER@.
7549 see also; cache_effective_group
7552 NAME: cache_effective_group
7555 DEFAULT_DOC: Use system group memberships of the cache_effective_user account
7556 LOC: Config.effectiveGroup
7558 Squid sets the GID to the effective user's default group ID
7559 (taken from the password file) and supplementary group list
7560 from the groups membership.
7562 If you want Squid to run with a specific GID regardless of
7563 the group memberships of the effective user then set this
7564 to the group (or GID) you want Squid to run as. When set
7565 all other group privileges of the effective user are ignored
7566 and only this GID is effective. If Squid is not started as
7567 root the user starting Squid MUST be member of the specified
7570 This option is not recommended by the Squid Team.
7571 Our preference is for administrators to configure a secure
7572 user account for squid with UID/GID matching system policies.
7575 NAME: httpd_suppress_version_string
7579 LOC: Config.onoff.httpd_suppress_version_string
7581 Do not send Squid version string in HTTP metadata and generated content
7582 such as HTML error pages. Squid version string is still present in certain
7583 SNMP responses, HTTP(S) Server response header field,
7584 various console output, and cache.log.
7587 NAME: visible_hostname
7589 LOC: Config.visibleHostname
7591 DEFAULT_DOC: Automatically detect the system host name
7593 If you want to present a special hostname in error messages, etc,
7594 define this. Otherwise, the return value of gethostname()
7595 will be used. If you have multiple caches in a cluster and
7596 get errors about IP-forwarding you must set them to have individual
7597 names with this setting.
7600 NAME: unique_hostname
7602 LOC: Config.uniqueHostname
7604 DEFAULT_DOC: Copy the value from visible_hostname
7606 If you want to have multiple machines with the same
7607 'visible_hostname' you must give each machine a different
7608 'unique_hostname' so forwarding loops can be detected.
7611 NAME: hostname_aliases
7613 LOC: Config.hostnameAliases
7616 A list of other DNS names your cache has.
7624 Minimum umask which should be enforced while the proxy
7625 is running, in addition to the umask set at startup.
7627 For a traditional octal representation of umasks, start
7632 HTTPD-ACCELERATOR OPTIONS
7633 -----------------------------------------------------------------------------
7636 NAME: httpd_accel_surrogate_id
7639 DEFAULT_DOC: visible_hostname is used if no specific ID is set.
7640 LOC: Config.Accel.surrogate_id
7642 Surrogates (http://www.esi.org/architecture_spec_1.0.html)
7643 need an identification token to allow control targeting. Because
7644 a farm of surrogates may all perform the same tasks, they may share
7645 an identification token.
7647 When the surrogate is a reverse-proxy, this ID is also
7648 used as cdn-id for CDN-Loop detection (RFC 8586).
7651 NAME: http_accel_surrogate_remote
7655 LOC: Config.onoff.surrogate_is_remote
7657 Remote surrogates (such as those in a CDN) honour the header
7658 "Surrogate-Control: no-store-remote".
7660 Set this to on to have squid behave as a remote surrogate.
7664 DELAY POOL PARAMETERS
7665 -----------------------------------------------------------------------------
7669 TYPE: delay_pool_count
7671 IFDEF: USE_DELAY_POOLS
7674 This represents the number of delay pools to be used. For example,
7675 if you have one class 2 delay pool and one class 3 delays pool, you
7676 have a total of 2 delay pools.
7678 See also delay_parameters, delay_class, delay_access for pool
7679 configuration details.
7683 TYPE: delay_pool_class
7685 IFDEF: USE_DELAY_POOLS
7688 This defines the class of each delay pool. There must be exactly one
7689 delay_class line for each delay pool. For example, to define two
7690 delay pools, one of class 2 and one of class 3, the settings above
7694 delay_pools 4 # 4 delay pools
7695 delay_class 1 2 # pool 1 is a class 2 pool
7696 delay_class 2 3 # pool 2 is a class 3 pool
7697 delay_class 3 4 # pool 3 is a class 4 pool
7698 delay_class 4 5 # pool 4 is a class 5 pool
7700 The delay pool classes are:
7702 class 1 Everything is limited by a single aggregate
7705 class 2 Everything is limited by a single aggregate
7706 bucket as well as an "individual" bucket chosen
7707 from bits 25 through 32 of the IPv4 address.
7709 class 3 Everything is limited by a single aggregate
7710 bucket as well as a "network" bucket chosen
7711 from bits 17 through 24 of the IP address and a
7712 "individual" bucket chosen from bits 17 through
7713 32 of the IPv4 address.
7715 class 4 Everything in a class 3 delay pool, with an
7716 additional limit on a per user basis. This
7717 only takes effect if the username is established
7718 in advance - by forcing authentication in your
7721 class 5 Requests are grouped according their tag (see
7722 external_acl's tag= reply).
7725 Each pool also requires a delay_parameters directive to configure the pool size
7726 and speed limits used whenever the pool is applied to a request. Along with
7727 a set of delay_access directives to determine when it is used.
7729 NOTE: If an IP address is a.b.c.d
7730 -> bits 25 through 32 are "d"
7731 -> bits 17 through 24 are "c"
7732 -> bits 17 through 32 are "c * 256 + d"
7734 NOTE-2: Due to the use of bitmasks in class 2,3,4 pools they only apply to
7735 IPv4 traffic. Class 1 and 5 pools may be used with IPv6 traffic.
7737 This clause only supports fast acl types.
7738 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
7740 See also delay_parameters and delay_access.
7744 TYPE: delay_pool_access
7746 DEFAULT_DOC: Deny using the pool, unless allow rules exist in squid.conf for the pool.
7747 IFDEF: USE_DELAY_POOLS
7750 This is used to determine which delay pool a request falls into.
7752 delay_access is sorted per pool and the matching starts with pool 1,
7753 then pool 2, ..., and finally pool N. The first delay pool where the
7754 request is allowed is selected for the request. If it does not allow
7755 the request to any pool then the request is not delayed (default).
7757 For example, if you want some_big_clients in delay
7758 pool 1 and lotsa_little_clients in delay pool 2:
7760 delay_access 1 allow some_big_clients
7761 delay_access 1 deny all
7762 delay_access 2 allow lotsa_little_clients
7763 delay_access 2 deny all
7764 delay_access 3 allow authenticated_clients
7766 See also delay_parameters and delay_class.
7770 NAME: delay_parameters
7771 TYPE: delay_pool_rates
7773 IFDEF: USE_DELAY_POOLS
7776 This defines the parameters for a delay pool. Each delay pool has
7777 a number of "buckets" associated with it, as explained in the
7778 description of delay_class.
7780 For a class 1 delay pool, the syntax is:
7782 delay_parameters pool aggregate
7784 For a class 2 delay pool:
7786 delay_parameters pool aggregate individual
7788 For a class 3 delay pool:
7790 delay_parameters pool aggregate network individual
7792 For a class 4 delay pool:
7794 delay_parameters pool aggregate network individual user
7796 For a class 5 delay pool:
7798 delay_parameters pool tagrate
7800 The option variables are:
7802 pool a pool number - ie, a number between 1 and the
7803 number specified in delay_pools as used in
7806 aggregate the speed limit parameters for the aggregate bucket
7809 individual the speed limit parameters for the individual
7810 buckets (class 2, 3).
7812 network the speed limit parameters for the network buckets
7815 user the speed limit parameters for the user buckets
7818 tagrate the speed limit parameters for the tag buckets
7821 A pair of delay parameters is written restore/maximum, where restore is
7822 the number of bytes (not bits - modem and network speeds are usually
7823 quoted in bits) per second placed into the bucket, and maximum is the
7824 maximum number of bytes which can be in the bucket at any time.
7826 There must be one delay_parameters line for each delay pool.
7829 For example, if delay pool number 1 is a class 2 delay pool as in the
7830 above example, and is being used to strictly limit each host to 64Kbit/sec
7831 (plus overheads), with no overall limit, the line is:
7833 delay_parameters 1 none 8000/8000
7835 Note that 8 x 8K Byte/sec -> 64K bit/sec.
7837 Note that the word 'none' is used to represent no limit.
7840 And, if delay pool number 2 is a class 3 delay pool as in the above
7841 example, and you want to limit it to a total of 256Kbit/sec (strict limit)
7842 with each 8-bit network permitted 64Kbit/sec (strict limit) and each
7843 individual host permitted 4800bit/sec with a bucket maximum size of 64Kbits
7844 to permit a decent web page to be downloaded at a decent speed
7845 (if the network is not being limited due to overuse) but slow down
7846 large downloads more significantly:
7848 delay_parameters 2 32000/32000 8000/8000 600/8000
7850 Note that 8 x 32K Byte/sec -> 256K bit/sec.
7851 8 x 8K Byte/sec -> 64K bit/sec.
7852 8 x 600 Byte/sec -> 4800 bit/sec.
7855 Finally, for a class 4 delay pool as in the example - each user will
7856 be limited to 128Kbits/sec no matter how many workstations they are logged into.:
7858 delay_parameters 4 32000/32000 8000/8000 600/64000 16000/16000
7861 See also delay_class and delay_access.
7865 NAME: delay_initial_bucket_level
7866 COMMENT: (percent, 0-100)
7869 IFDEF: USE_DELAY_POOLS
7870 LOC: Config.Delay.initial
7872 The initial bucket percentage is used to determine how much is put
7873 in each bucket when squid starts, is reconfigured, or first notices
7874 a host accessing it (in class 2 and class 3, individual hosts and
7875 networks only have buckets associated with them once they have been
7880 CLIENT DELAY POOL PARAMETERS
7881 -----------------------------------------------------------------------------
7884 NAME: client_delay_pools
7885 TYPE: client_delay_pool_count
7887 IFDEF: USE_DELAY_POOLS
7888 LOC: Config.ClientDelay
7890 This option specifies the number of client delay pools used. It must
7891 preceed other client_delay_* options.
7894 client_delay_pools 2
7896 See also client_delay_parameters and client_delay_access.
7899 NAME: client_delay_initial_bucket_level
7900 COMMENT: (percent, 0-no_limit)
7903 IFDEF: USE_DELAY_POOLS
7904 LOC: Config.ClientDelay.initial
7906 This option determines the initial bucket size as a percentage of
7907 max_bucket_size from client_delay_parameters. Buckets are created
7908 at the time of the "first" connection from the matching IP. Idle
7909 buckets are periodically deleted up.
7911 You can specify more than 100 percent but note that such "oversized"
7912 buckets are not refilled until their size goes down to max_bucket_size
7913 from client_delay_parameters.
7916 client_delay_initial_bucket_level 50
7919 NAME: client_delay_parameters
7920 TYPE: client_delay_pool_rates
7922 IFDEF: USE_DELAY_POOLS
7923 LOC: Config.ClientDelay
7926 This option configures client-side bandwidth limits using the
7929 client_delay_parameters pool speed_limit max_bucket_size
7931 pool is an integer ID used for client_delay_access matching.
7933 speed_limit is bytes added to the bucket per second.
7935 max_bucket_size is the maximum size of a bucket, enforced after any
7936 speed_limit additions.
7938 Please see the delay_parameters option for more information and
7942 client_delay_parameters 1 1024 2048
7943 client_delay_parameters 2 51200 16384
7945 See also client_delay_access.
7949 NAME: client_delay_access
7950 TYPE: client_delay_pool_access
7952 DEFAULT_DOC: Deny use of the pool, unless allow rules exist in squid.conf for the pool.
7953 IFDEF: USE_DELAY_POOLS
7954 LOC: Config.ClientDelay
7956 This option determines the client-side delay pool for the
7959 client_delay_access pool_ID allow|deny acl_name
7961 All client_delay_access options are checked in their pool ID
7962 order, starting with pool 1. The first checked pool with allowed
7963 request is selected for the request. If no ACL matches or there
7964 are no client_delay_access options, the request bandwidth is not
7967 The ACL-selected pool is then used to find the
7968 client_delay_parameters for the request. Client-side pools are
7969 not used to aggregate clients. Clients are always aggregated
7970 based on their source IP addresses (one bucket per source IP).
7972 This clause only supports fast acl types.
7973 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
7974 Additionally, only the client TCP connection details are available.
7975 ACLs testing HTTP properties will not work.
7977 Please see delay_access for more examples.
7980 client_delay_access 1 allow low_rate_network
7981 client_delay_access 2 allow vips_network
7984 See also client_delay_parameters and client_delay_pools.
7987 NAME: response_delay_pool
7988 TYPE: response_delay_pool_parameters
7990 IFDEF: USE_DELAY_POOLS
7991 LOC: Config.MessageDelay
7993 This option configures client response bandwidth limits using the
7996 response_delay_pool name [option=value] ...
7998 name the response delay pool name
8002 individual-restore The speed limit of an individual
8003 bucket(bytes/s). To be used in conjunction
8004 with 'individual-maximum'.
8006 individual-maximum The maximum number of bytes which can
8007 be placed into the individual bucket. To be used
8008 in conjunction with 'individual-restore'.
8010 aggregate-restore The speed limit for the aggregate
8011 bucket(bytes/s). To be used in conjunction with
8012 'aggregate-maximum'.
8014 aggregate-maximum The maximum number of bytes which can
8015 be placed into the aggregate bucket. To be used
8016 in conjunction with 'aggregate-restore'.
8018 initial-bucket-level The initial bucket size as a percentage
8019 of individual-maximum.
8021 Individual and(or) aggregate bucket options may not be specified,
8022 meaning no individual and(or) aggregate speed limitation.
8023 See also response_delay_pool_access and delay_parameters for
8024 terminology details.
8027 NAME: response_delay_pool_access
8028 TYPE: response_delay_pool_access
8030 DEFAULT_DOC: Deny use of the pool, unless allow rules exist in squid.conf for the pool.
8031 IFDEF: USE_DELAY_POOLS
8032 LOC: Config.MessageDelay
8034 Determines whether a specific named response delay pool is used
8035 for the transaction. The syntax for this directive is:
8037 response_delay_pool_access pool_name allow|deny acl_name
8039 All response_delay_pool_access options are checked in the order
8040 they appear in this configuration file. The first rule with a
8041 matching ACL wins. If (and only if) an "allow" rule won, Squid
8042 assigns the response to the corresponding named delay pool.
8046 WCCPv1 AND WCCPv2 CONFIGURATION OPTIONS
8047 -----------------------------------------------------------------------------
8052 LOC: Config.Wccp.router
8054 DEFAULT_DOC: WCCP disabled.
8057 Use this option to define your WCCP ``home'' router for
8060 wccp_router supports a single WCCP(v1) router
8062 wccp2_router supports multiple WCCPv2 routers
8064 only one of the two may be used at the same time and defines
8065 which version of WCCP to use.
8069 TYPE: IpAddress_list
8070 LOC: Config.Wccp2.router
8072 DEFAULT_DOC: WCCPv2 disabled.
8075 Use this option to define your WCCP ``home'' router for
8078 wccp_router supports a single WCCP(v1) router
8080 wccp2_router supports multiple WCCPv2 routers
8082 only one of the two may be used at the same time and defines
8083 which version of WCCP to use.
8088 LOC: Config.Wccp.version
8092 This directive is only relevant if you need to set up WCCP(v1)
8093 to some very old and end-of-life Cisco routers. In all other
8094 setups it must be left unset or at the default setting.
8095 It defines an internal version in the WCCP(v1) protocol,
8096 with version 4 being the officially documented protocol.
8098 According to some users, Cisco IOS 11.2 and earlier only
8099 support WCCP version 3. If you're using that or an earlier
8100 version of IOS, you may need to change this value to 3, otherwise
8101 do not specify this parameter.
8104 NAME: wccp2_rebuild_wait
8106 LOC: Config.Wccp2.rebuildwait
8110 If this is enabled Squid will wait for the cache dir rebuild to finish
8111 before sending the first wccp2 HereIAm packet
8114 NAME: wccp2_forwarding_method
8116 LOC: Config.Wccp2.forwarding_method
8120 WCCP2 allows the setting of forwarding methods between the
8121 router/switch and the cache. Valid values are as follows:
8123 gre - GRE encapsulation (forward the packet in a GRE/WCCP tunnel)
8124 l2 - L2 redirect (forward the packet using Layer 2/MAC rewriting)
8126 Currently (as of IOS 12.4) cisco routers only support GRE.
8127 Cisco switches only support the L2 redirect assignment method.
8130 NAME: wccp2_return_method
8132 LOC: Config.Wccp2.return_method
8136 WCCP2 allows the setting of return methods between the
8137 router/switch and the cache for packets that the cache
8138 decides not to handle. Valid values are as follows:
8140 gre - GRE encapsulation (forward the packet in a GRE/WCCP tunnel)
8141 l2 - L2 redirect (forward the packet using Layer 2/MAC rewriting)
8143 Currently (as of IOS 12.4) cisco routers only support GRE.
8144 Cisco switches only support the L2 redirect assignment.
8146 If the "ip wccp redirect exclude in" command has been
8147 enabled on the cache interface, then it is still safe for
8148 the proxy server to use a l2 redirect method even if this
8149 option is set to GRE.
8152 NAME: wccp2_assignment_method
8154 LOC: Config.Wccp2.assignment_method
8158 WCCP2 allows the setting of methods to assign the WCCP hash
8159 Valid values are as follows:
8161 hash - Hash assignment
8162 mask - Mask assignment
8164 As a general rule, cisco routers support the hash assignment method
8165 and cisco switches support the mask assignment method.
8170 LOC: Config.Wccp2.info
8171 DEFAULT_IF_NONE: standard 0
8172 DEFAULT_DOC: Use the 'web-cache' standard service.
8175 WCCP2 allows for multiple traffic services. There are two
8176 types: "standard" and "dynamic". The standard type defines
8177 one service id - http (id 0). The dynamic service ids can be from
8178 51 to 255 inclusive. In order to use a dynamic service id
8179 one must define the type of traffic to be redirected; this is done
8180 using the wccp2_service_info option.
8182 The "standard" type does not require a wccp2_service_info option,
8183 just specifying the service id will suffice.
8185 MD5 service authentication can be enabled by adding
8186 "password=<password>" to the end of this service declaration.
8190 wccp2_service standard 0 # for the 'web-cache' standard service
8191 wccp2_service dynamic 80 # a dynamic service type which will be
8192 # fleshed out with subsequent options.
8193 wccp2_service standard 0 password=foo
8196 NAME: wccp2_service_info
8197 TYPE: wccp2_service_info
8198 LOC: Config.Wccp2.info
8202 Dynamic WCCPv2 services require further information to define the
8203 traffic you wish to have diverted.
8207 wccp2_service_info <id> protocol=<protocol> flags=<flag>,<flag>..
8208 priority=<priority> ports=<port>,<port>..
8210 The relevant WCCPv2 flags:
8211 + src_ip_hash, dst_ip_hash
8212 + source_port_hash, dst_port_hash
8213 + src_ip_alt_hash, dst_ip_alt_hash
8214 + src_port_alt_hash, dst_port_alt_hash
8217 The port list can be one to eight entries.
8221 wccp2_service_info 80 protocol=tcp flags=src_ip_hash,ports_source
8222 priority=240 ports=80
8224 Note: the service id must have been defined by a previous
8225 'wccp2_service dynamic <id>' entry.
8230 LOC: Config.Wccp2.weight
8234 Each cache server gets assigned a set of the destination
8235 hash proportional to their weight.
8240 LOC: Config.Wccp.address
8242 DEFAULT_DOC: Address selected by the operating system.
8245 Use this option if you require WCCP(v1) to use a specific
8248 The default behavior is to not bind to any specific address.
8253 LOC: Config.Wccp2.address
8255 DEFAULT_DOC: Address selected by the operating system.
8258 Use this option if you require WCCPv2 to use a specific
8261 The default behavior is to not bind to any specific address.
8265 PERSISTENT CONNECTION HANDLING
8266 -----------------------------------------------------------------------------
8268 Also see "pconn_timeout" in the TIMEOUTS section
8271 NAME: client_persistent_connections
8273 LOC: Config.onoff.client_pconns
8276 Persistent connection support for clients.
8277 Squid uses persistent connections (when allowed). You can use
8278 this option to disable persistent connections with clients.
8281 NAME: server_persistent_connections
8283 LOC: Config.onoff.server_pconns
8286 Persistent connection support for servers.
8287 Squid uses persistent connections (when allowed). You can use
8288 this option to disable persistent connections with servers.
8291 NAME: persistent_connection_after_error
8293 LOC: Config.onoff.error_pconns
8296 With this directive the use of persistent connections after
8297 HTTP errors can be disabled. Useful if you have clients
8298 who fail to handle errors on persistent connections proper.
8301 NAME: detect_broken_pconn
8303 LOC: Config.onoff.detect_broken_server_pconns
8306 Some servers have been found to incorrectly signal the use
8307 of HTTP/1.0 persistent connections even on replies not
8308 compatible, causing significant delays. This server problem
8309 has mostly been seen on redirects.
8311 By enabling this directive Squid attempts to detect such
8312 broken replies and automatically assume the reply is finished
8313 after 10 seconds timeout.
8317 CACHE DIGEST OPTIONS
8318 -----------------------------------------------------------------------------
8321 NAME: digest_generation
8322 IFDEF: USE_CACHE_DIGESTS
8324 LOC: Config.onoff.digest_generation
8327 This controls whether the server will generate a Cache Digest
8328 of its contents. By default, Cache Digest generation is
8329 enabled if Squid is compiled with --enable-cache-digests defined.
8332 NAME: digest_bits_per_entry
8333 IFDEF: USE_CACHE_DIGESTS
8335 LOC: Config.digest.bits_per_entry
8338 This is the number of bits of the server's Cache Digest which
8339 will be associated with the Digest entry for a given HTTP
8340 Method and URL (public key) combination. The default is 5.
8343 NAME: digest_rebuild_period
8344 IFDEF: USE_CACHE_DIGESTS
8347 LOC: Config.digest.rebuild_period
8350 This is the wait time between Cache Digest rebuilds.
8353 NAME: digest_rewrite_period
8355 IFDEF: USE_CACHE_DIGESTS
8357 LOC: Config.digest.rewrite_period
8360 This is the wait time between Cache Digest writes to
8364 NAME: digest_swapout_chunk_size
8367 IFDEF: USE_CACHE_DIGESTS
8368 LOC: Config.digest.swapout_chunk_size
8371 This is the number of bytes of the Cache Digest to write to
8372 disk at a time. It defaults to 4096 bytes (4KB), the Squid
8376 NAME: digest_rebuild_chunk_percentage
8377 COMMENT: (percent, 0-100)
8378 IFDEF: USE_CACHE_DIGESTS
8380 LOC: Config.digest.rebuild_chunk_percentage
8383 This is the percentage of the Cache Digest to be scanned at a
8384 time. By default it is set to 10% of the Cache Digest.
8389 -----------------------------------------------------------------------------
8394 LOC: Config.Port.snmp
8396 DEFAULT_DOC: SNMP disabled.
8399 The port number where Squid listens for SNMP requests. To enable
8400 SNMP support set this to a suitable port number. Port number
8401 3401 is often used for the Squid SNMP agent. By default it's
8402 set to "0" (disabled)
8410 LOC: Config.accessList.snmp
8412 DEFAULT_DOC: Deny, unless rules exist in squid.conf.
8415 Allowing or denying access to the SNMP port.
8417 All access to the agent is denied by default.
8420 snmp_access allow|deny [!]aclname ...
8422 This clause only supports fast acl types.
8423 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
8426 snmp_access allow snmppublic localhost
8427 snmp_access deny all
8430 NAME: snmp_incoming_address
8432 LOC: Config.Addrs.snmp_incoming
8434 DEFAULT_DOC: Accept SNMP packets from all machine interfaces.
8437 Just like 'udp_incoming_address', but for the SNMP port.
8439 snmp_incoming_address is used for the SNMP socket receiving
8440 messages from SNMP agents.
8442 The default snmp_incoming_address is to listen on all
8443 available network interfaces.
8446 NAME: snmp_outgoing_address
8448 LOC: Config.Addrs.snmp_outgoing
8450 DEFAULT_DOC: Use snmp_incoming_address or an address selected by the operating system.
8453 Just like 'udp_outgoing_address', but for the SNMP port.
8455 snmp_outgoing_address is used for SNMP packets returned to SNMP
8458 If snmp_outgoing_address is not set it will use the same socket
8459 as snmp_incoming_address. Only change this if you want to have
8460 SNMP replies sent using another address than where this Squid
8461 listens for SNMP queries.
8463 NOTE, snmp_incoming_address and snmp_outgoing_address can not have
8464 the same value since they both use the same port.
8469 -----------------------------------------------------------------------------
8472 NAME: icp_port udp_port
8475 DEFAULT_DOC: ICP disabled.
8476 LOC: Config.Port.icp
8478 The port number where Squid sends and receives ICP queries to
8479 and from neighbor caches. The standard UDP port for ICP is 3130.
8482 icp_port @DEFAULT_ICP_PORT@
8489 DEFAULT_DOC: HTCP disabled.
8490 LOC: Config.Port.htcp
8492 The port number where Squid sends and receives HTCP queries to
8493 and from neighbor caches. To turn it on you want to set it to
8500 NAME: log_icp_queries
8504 LOC: Config.onoff.log_udp
8506 If set, ICP queries are logged to access.log. You may wish
8507 do disable this if your ICP load is VERY high to speed things
8508 up or to simplify log analysis.
8511 NAME: udp_incoming_address
8513 LOC:Config.Addrs.udp_incoming
8515 DEFAULT_DOC: Accept packets from all machine interfaces.
8517 udp_incoming_address is used for UDP packets received from other
8520 The default behavior is to not bind to any specific address.
8522 Only change this if you want to have all UDP queries received on
8523 a specific interface/address.
8525 NOTE: udp_incoming_address is used by the ICP, HTCP, and DNS
8526 modules. Altering it will affect all of them in the same manner.
8528 see also; udp_outgoing_address
8530 NOTE, udp_incoming_address and udp_outgoing_address can not
8531 have the same value since they both use the same port.
8534 NAME: udp_outgoing_address
8536 LOC: Config.Addrs.udp_outgoing
8538 DEFAULT_DOC: Use udp_incoming_address or an address selected by the operating system.
8540 udp_outgoing_address is used for UDP packets sent out to other
8543 The default behavior is to not bind to any specific address.
8545 Instead it will use the same socket as udp_incoming_address.
8546 Only change this if you want to have UDP queries sent using another
8547 address than where this Squid listens for UDP queries from other
8550 NOTE: udp_outgoing_address is used by the ICP, HTCP, and DNS
8551 modules. Altering it will affect all of them in the same manner.
8553 see also; udp_incoming_address
8555 NOTE, udp_incoming_address and udp_outgoing_address can not
8556 have the same value since they both use the same port.
8563 LOC: Config.onoff.icp_hit_stale
8565 If you want to return ICP_HIT for stale cache objects, set this
8566 option to 'on'. If you have sibling relationships with caches
8567 in other administrative domains, this should be 'off'. If you only
8568 have sibling relationships with caches under your control,
8569 it is probably okay to set this to 'on'.
8570 If set to 'on', your siblings should use the option "allow-miss"
8571 on their cache_peer lines for connecting to you.
8574 NAME: minimum_direct_hops
8577 LOC: Config.minDirectHops
8579 If using the ICMP pinging stuff, do direct fetches for sites
8580 which are no more than this many hops away.
8583 NAME: minimum_direct_rtt
8587 LOC: Config.minDirectRtt
8589 If using the ICMP pinging stuff, do direct fetches for sites
8590 which are no more than this many rtt milliseconds away.
8596 LOC: Config.Netdb.low
8598 The low water mark for the ICMP measurement database.
8600 Note: high watermark controlled by netdb_high directive.
8602 These watermarks are counts, not percents. The defaults are
8603 (low) 900 and (high) 1000. When the high water mark is
8604 reached, database entries will be deleted until the low
8611 LOC: Config.Netdb.high
8613 The high water mark for the ICMP measurement database.
8615 Note: low watermark controlled by netdb_low directive.
8617 These watermarks are counts, not percents. The defaults are
8618 (low) 900 and (high) 1000. When the high water mark is
8619 reached, database entries will be deleted until the low
8623 NAME: netdb_ping_period
8625 LOC: Config.Netdb.period
8628 The minimum period for measuring a site. There will be at
8629 least this much delay between successive pings to the same
8630 network. The default is five minutes.
8637 LOC: Config.onoff.query_icmp
8639 If you want to ask your peers to include ICMP data in their ICP
8640 replies, enable this option.
8642 If your peer has configured Squid (during compilation) with
8643 '--enable-icmp' that peer will send ICMP pings to origin server
8644 sites of the URLs it receives. If you enable this option the
8645 ICP replies from that peer will include the ICMP data (if available).
8646 Then, when choosing a parent cache, Squid will choose the parent with
8647 the minimal RTT to the origin server. When this happens, the
8648 hierarchy field of the access.log will be
8649 "CLOSEST_PARENT_MISS". This option is off by default.
8652 NAME: test_reachability
8656 LOC: Config.onoff.test_reachability
8658 When this is 'on', ICP MISS replies will be ICP_MISS_NOFETCH
8659 instead of ICP_MISS if the target host is NOT in the ICMP
8660 database, or has a zero RTT.
8663 NAME: icp_query_timeout
8666 DEFAULT_DOC: Dynamic detection.
8668 LOC: Config.Timeout.icp_query
8670 Normally Squid will automatically determine an optimal ICP
8671 query timeout value based on the round-trip-time of recent ICP
8672 queries. If you want to override the value determined by
8673 Squid, set this 'icp_query_timeout' to a non-zero value. This
8674 value is specified in MILLISECONDS, so, to use a 2-second
8675 timeout (the old default), you would write:
8677 icp_query_timeout 2000
8680 NAME: maximum_icp_query_timeout
8684 LOC: Config.Timeout.icp_query_max
8686 Normally the ICP query timeout is determined dynamically. But
8687 sometimes it can lead to very large values (say 5 seconds).
8688 Use this option to put an upper limit on the dynamic timeout
8689 value. Do NOT use this option to always use a fixed (instead
8690 of a dynamic) timeout value. To set a fixed timeout see the
8691 'icp_query_timeout' directive.
8694 NAME: minimum_icp_query_timeout
8698 LOC: Config.Timeout.icp_query_min
8700 Normally the ICP query timeout is determined dynamically. But
8701 sometimes it can lead to very small timeouts, even lower than
8702 the normal latency variance on your link due to traffic.
8703 Use this option to put an lower limit on the dynamic timeout
8704 value. Do NOT use this option to always use a fixed (instead
8705 of a dynamic) timeout value. To set a fixed timeout see the
8706 'icp_query_timeout' directive.
8709 NAME: background_ping_rate
8713 LOC: Config.backgroundPingRate
8715 Controls how often the ICP pings are sent to siblings that
8716 have background-ping set.
8720 MULTICAST ICP OPTIONS
8721 -----------------------------------------------------------------------------
8726 LOC: Config.mcast_group_list
8729 This tag specifies a list of multicast groups which your server
8730 should join to receive multicasted ICP queries.
8732 NOTE! Be very careful what you put here! Be sure you
8733 understand the difference between an ICP _query_ and an ICP
8734 _reply_. This option is to be set only if you want to RECEIVE
8735 multicast queries. Do NOT set this option to SEND multicast
8736 ICP (use cache_peer for that). ICP replies are always sent via
8737 unicast, so this option does not affect whether or not you will
8738 receive replies from multicast group members.
8740 You must be very careful to NOT use a multicast address which
8741 is already in use by another group of caches.
8743 If you are unsure about multicast, please read the Multicast
8744 chapter in the Squid FAQ (http://www.squid-cache.org/FAQ/).
8746 Usage: mcast_groups 239.128.16.128 224.0.1.20
8748 By default, Squid doesn't listen on any multicast groups.
8751 NAME: mcast_icp_query_timeout
8755 LOC: Config.Timeout.mcast_icp_query
8757 For multicast peers, Squid regularly sends out ICP "probes" to
8758 count how many other peers are listening on the given multicast
8759 address. This value specifies how long Squid should wait to
8760 count all the replies. The default is 2000 msec, or 2
8765 INTERNAL ICON OPTIONS
8766 -----------------------------------------------------------------------------
8769 NAME: icon_directory
8771 LOC: Config.icons.directory
8772 DEFAULT: @DEFAULT_ICON_DIR@
8774 Where the icons are stored. These are normally kept in
8778 NAME: global_internal_static
8780 LOC: Config.onoff.global_internal_static
8783 This directive controls is Squid should intercept all requests for
8784 /squid-internal-static/ no matter which host the URL is requesting
8785 (default on setting), or if nothing special should be done for
8786 such URLs (off setting). The purpose of this directive is to make
8787 icons etc work better in complex cache hierarchies where it may
8788 not always be possible for all corners in the cache mesh to reach
8789 the server generating a directory listing.
8792 NAME: short_icon_urls
8794 LOC: Config.icons.use_short_names
8797 If this is enabled Squid will use short URLs for icons.
8798 If disabled it will revert to the old behavior of including
8799 it's own name and port in the URL.
8801 If you run a complex cache hierarchy with a mix of Squid and
8802 other proxies you may need to disable this directive.
8807 -----------------------------------------------------------------------------
8810 NAME: error_directory
8812 LOC: Config.errorDirectory
8814 DEFAULT_DOC: Send error pages in the clients preferred language
8816 If you wish to create your own versions of the default
8817 error files to customize them to suit your company copy
8818 the error/template files to another directory and point
8821 WARNING: This option will disable multi-language support
8822 on error pages if used.
8824 The squid developers are interested in making squid available in
8825 a wide variety of languages. If you are making translations for a
8826 language that Squid does not currently provide please consider
8827 contributing your translation back to the project.
8828 https://wiki.squid-cache.org/Translations
8830 The squid developers working on translations are happy to supply drop-in
8831 translated error files in exchange for any new language contributions.
8834 NAME: error_default_language
8835 IFDEF: USE_ERR_LOCALES
8837 LOC: Config.errorDefaultLanguage
8839 DEFAULT_DOC: Generate English language pages.
8841 Set the default language which squid will send error pages in
8842 if no existing translation matches the clients language
8845 If unset (default) generic English will be used.
8847 The squid developers are interested in making squid available in
8848 a wide variety of languages. If you are interested in making
8849 translations for any language see the squid wiki for details.
8850 https://wiki.squid-cache.org/Translations
8853 NAME: error_log_languages
8854 IFDEF: USE_ERR_LOCALES
8856 LOC: Config.errorLogMissingLanguages
8859 Log to cache.log what languages users are attempting to
8860 auto-negotiate for translations.
8862 Successful negotiations are not logged. Only failures
8863 have meaning to indicate that Squid may need an upgrade
8864 of its error page translations.
8867 NAME: err_page_stylesheet
8869 LOC: Config.errorStylesheet
8870 DEFAULT: @DEFAULT_CONFIG_DIR@/errorpage.css
8872 CSS Stylesheet to pattern the display of Squid default error pages.
8874 For information on CSS see http://www.w3.org/Style/CSS/
8879 LOC: Config.errHtmlText
8882 HTML text to include in error messages. Make this a "mailto"
8883 URL to your admin address, or maybe just a link to your
8884 organizations Web page.
8886 To include this in your error messages, you must rewrite
8887 the error template files (found in the "errors" directory).
8888 Wherever you want the 'err_html_text' line to appear,
8889 insert a %L tag in the error template file.
8892 NAME: email_err_data
8895 LOC: Config.onoff.emailErrData
8898 If enabled, information about the occurred error will be
8899 included in the mailto links of the ERR pages (if %W is set)
8900 so that the email body contains the data.
8901 Syntax is <A HREF="mailto:%w%W">%w</A>
8906 LOC: Config.denyInfoList
8909 Usage: deny_info err_page_name acl
8910 or deny_info http://... acl
8911 or deny_info TCP_RESET acl
8913 This can be used to return a ERR_ page for requests which
8914 do not pass the 'http_access' rules. Squid remembers the last
8915 acl it evaluated in http_access, and if a 'deny_info' line exists
8916 for that ACL Squid returns a corresponding error page.
8918 The acl is typically the last acl on the http_access deny line which
8919 denied access. The exceptions to this rule are:
8920 - When Squid needs to request authentication credentials. It's then
8921 the first authentication related acl encountered
8922 - When none of the http_access lines matches. It's then the last
8923 acl processed on the last http_access line.
8924 - When the decision to deny access was made by an adaptation service,
8925 the acl name is the corresponding eCAP or ICAP service_name.
8927 NP: If providing your own custom error pages with error_directory
8928 you may also specify them by your custom file name:
8929 Example: deny_info ERR_CUSTOM_ACCESS_DENIED bad_guys
8931 By default Squid will send "403 Forbidden". A different 4xx or 5xx
8932 may be specified by prefixing the file name with the code and a colon.
8933 e.g. 404:ERR_CUSTOM_ACCESS_DENIED
8935 Alternatively you can tell Squid to reset the TCP connection
8936 by specifying TCP_RESET.
8938 Or you can specify an error URL or URL pattern. The browsers will
8939 get redirected to the specified URL after formatting tags have
8940 been replaced. Redirect will be done with 302 or 307 according to
8941 HTTP/1.1 specs. A different 3xx code may be specified by prefixing
8942 the URL. e.g. 303:http://example.com/
8945 %a - username (if available. Password NOT included)
8946 %A - Local listening IP address the client connection was connected to
8949 %E - Error description
8951 %H - Request domain name
8952 %i - Client IP Address
8954 %O - Unescaped message result from external ACL helper
8955 %o - Message result from external ACL helper
8956 %p - Request Port number
8957 %P - Request Protocol name
8958 %R - Request URL path
8959 %T - Timestamp in RFC 1123 format
8960 %U - Full canonical URL from client
8961 (HTTPS URLs terminate with *)
8962 %u - Full canonical URL from client
8963 %w - Admin email from squid.conf
8965 %% - Literal percent (%) code
8970 OPTIONS INFLUENCING REQUEST FORWARDING
8971 -----------------------------------------------------------------------------
8974 NAME: nonhierarchical_direct
8976 LOC: Config.onoff.nonhierarchical_direct
8979 By default, Squid will send any non-hierarchical requests
8980 (not cacheable request type) direct to origin servers.
8982 When this is set to "off", Squid will prefer to send these
8983 requests to parents.
8985 Note that in most configurations, by turning this off you will only
8986 add latency to these request without any improvement in global hit
8989 This option only sets a preference. If the parent is unavailable a
8990 direct connection to the origin server may still be attempted. To
8991 completely prevent direct connections use never_direct.
8996 LOC: Config.onoff.prefer_direct
8999 Normally Squid tries to use parents for most requests. If you for some
9000 reason like it to first try going direct and only use a parent if
9001 going direct fails set this to on.
9003 By combining nonhierarchical_direct off and prefer_direct on you
9004 can set up Squid to use a parent as a backup path if going direct
9007 Note: If you want Squid to use parents for all requests see
9008 the never_direct directive. prefer_direct only modifies how Squid
9009 acts on cacheable requests.
9012 NAME: cache_miss_revalidate
9016 LOC: Config.onoff.cache_miss_revalidate
9018 RFC 7232 defines a conditional request mechanism to prevent
9019 response objects being unnecessarily transferred over the network.
9020 If that mechanism is used by the client and a cache MISS occurs
9021 it can prevent new cache entries being created.
9023 This option determines whether Squid on cache MISS will pass the
9024 client revalidation request to the server or tries to fetch new
9025 content for caching. It can be useful while the cache is mostly
9026 empty to more quickly have the cache populated by generating
9027 non-conditional GETs.
9029 When set to 'on' (default), Squid will pass all client If-* headers
9030 to the server. This permits server responses without a cacheable
9031 payload to be delivered and on MISS no new cache entry is created.
9033 When set to 'off' and if the request is cacheable, Squid will
9034 remove the clients If-Modified-Since and If-None-Match headers from
9035 the request sent to the server. This requests a 200 status response
9036 from the server to create a new cache entry with.
9041 LOC: Config.accessList.AlwaysDirect
9043 DEFAULT_DOC: Prevent any cache_peer being used for this request.
9045 Usage: always_direct allow|deny [!]aclname ...
9047 Here you can use ACL elements to specify requests which should
9048 ALWAYS be forwarded by Squid to the origin servers without using
9049 any peers. For example, to always directly forward requests for
9050 local servers ignoring any parents or siblings you may have use
9053 acl local-servers dstdomain my.domain.net
9054 always_direct allow local-servers
9056 To always forward FTP requests directly, use
9059 always_direct allow FTP
9061 NOTE: There is a similar, but opposite option named
9062 'never_direct'. You need to be aware that "always_direct deny
9063 foo" is NOT the same thing as "never_direct allow foo". You
9064 may need to use a deny rule to exclude a more-specific case of
9065 some other rule. Example:
9067 acl local-external dstdomain external.foo.net
9068 acl local-servers dstdomain .foo.net
9069 always_direct deny local-external
9070 always_direct allow local-servers
9072 NOTE: If your goal is to make the client forward the request
9073 directly to the origin server bypassing Squid then this needs
9074 to be done in the client configuration. Squid configuration
9075 can only tell Squid how Squid should fetch the object.
9077 NOTE: This directive is not related to caching. The replies
9078 is cached as usual even if you use always_direct. To not cache
9079 the replies see the 'cache' directive.
9081 This clause supports both fast and slow acl types.
9082 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
9087 LOC: Config.accessList.NeverDirect
9089 DEFAULT_DOC: Allow DNS results to be used for this request.
9091 Usage: never_direct allow|deny [!]aclname ...
9093 never_direct is the opposite of always_direct. Please read
9094 the description for always_direct if you have not already.
9096 With 'never_direct' you can use ACL elements to specify
9097 requests which should NEVER be forwarded directly to origin
9098 servers. For example, to force the use of a proxy for all
9099 requests, except those in your local domain use something like:
9101 acl local-servers dstdomain .foo.net
9102 never_direct deny local-servers
9103 never_direct allow all
9105 or if Squid is inside a firewall and there are local intranet
9106 servers inside the firewall use something like:
9108 acl local-intranet dstdomain .foo.net
9109 acl local-external dstdomain external.foo.net
9110 always_direct deny local-external
9111 always_direct allow local-intranet
9112 never_direct allow all
9114 This clause supports both fast and slow acl types.
9115 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
9119 ADVANCED NETWORKING OPTIONS
9120 -----------------------------------------------------------------------------
9123 NAME: incoming_udp_average incoming_icp_average
9126 LOC: Config.comm_incoming.udp.average
9128 Heavy voodoo here. I can't even believe you are reading this.
9129 Are you crazy? Don't even think about adjusting these unless
9130 you understand the algorithms in comm_select.c first!
9133 NAME: incoming_tcp_average incoming_http_average
9136 LOC: Config.comm_incoming.tcp.average
9138 Heavy voodoo here. I can't even believe you are reading this.
9139 Are you crazy? Don't even think about adjusting these unless
9140 you understand the algorithms in comm_select.c first!
9143 NAME: incoming_dns_average
9146 LOC: Config.comm_incoming.dns.average
9148 Heavy voodoo here. I can't even believe you are reading this.
9149 Are you crazy? Don't even think about adjusting these unless
9150 you understand the algorithms in comm_select.c first!
9153 NAME: min_udp_poll_cnt min_icp_poll_cnt
9156 LOC: Config.comm_incoming.udp.min_poll
9158 Heavy voodoo here. I can't even believe you are reading this.
9159 Are you crazy? Don't even think about adjusting these unless
9160 you understand the algorithms in comm_select.c first!
9163 NAME: min_dns_poll_cnt
9166 LOC: Config.comm_incoming.dns.min_poll
9168 Heavy voodoo here. I can't even believe you are reading this.
9169 Are you crazy? Don't even think about adjusting these unless
9170 you understand the algorithms in comm_select.c first!
9173 NAME: min_tcp_poll_cnt min_http_poll_cnt
9176 LOC: Config.comm_incoming.tcp.min_poll
9178 Heavy voodoo here. I can't even believe you are reading this.
9179 Are you crazy? Don't even think about adjusting these unless
9180 you understand the algorithms in comm_select.c first!
9186 LOC: Config.accept_filter
9190 The name of an accept(2) filter to install on Squid's
9191 listen socket(s). This feature is perhaps specific to
9192 FreeBSD and requires support in the kernel.
9194 The 'httpready' filter delays delivering new connections
9195 to Squid until a full HTTP request has been received.
9196 See the accf_http(9) man page for details.
9198 The 'dataready' filter delays delivering new connections
9199 to Squid until there is some data to process.
9200 See the accf_dataready(9) man page for details.
9204 The 'data' filter delays delivering of new connections
9205 to Squid until there is some data to process by TCP_ACCEPT_DEFER.
9206 You may optionally specify a number of seconds to wait by
9207 'data=N' where N is the number of seconds. Defaults to 30
9208 if not specified. See the tcp(7) man page for details.
9211 accept_filter httpready
9216 NAME: client_ip_max_connections
9218 LOC: Config.client_ip_max_connections
9220 DEFAULT_DOC: No limit.
9222 Set an absolute limit on the number of connections a single
9223 client IP can use. Any more than this and Squid will begin to drop
9224 new connections from the client until it closes some links.
9226 Note that this is a global limit. It affects all HTTP, HTCP, and FTP
9227 connections from the client. For finer control use the ACL access controls.
9229 Requires client_db to be enabled (the default).
9231 WARNING: This may noticeably slow down traffic received via external proxies
9232 or NAT devices and cause them to rebound error messages back to their clients.
9235 NAME: tcp_recv_bufsize
9239 DEFAULT_DOC: Use operating system TCP defaults.
9240 LOC: Config.tcpRcvBufsz
9242 Size of receive buffer to set for TCP sockets. Probably just
9243 as easy to change your kernel's default.
9244 Omit from squid.conf to use the default buffer size.
9249 -----------------------------------------------------------------------------
9256 LOC: Adaptation::Icap::TheConfig.onoff
9259 If you want to enable the ICAP module support, set this to on.
9262 NAME: icap_connect_timeout
9265 LOC: Adaptation::Icap::TheConfig.connect_timeout_raw
9268 This parameter specifies how long to wait for the TCP connect to
9269 the requested ICAP server to complete before giving up and either
9270 terminating the HTTP transaction or bypassing the failure.
9272 The default for optional services is peer_connect_timeout.
9273 The default for essential services is connect_timeout.
9274 If this option is explicitly set, its value applies to all services.
9277 NAME: icap_io_timeout
9281 DEFAULT_DOC: Use read_timeout.
9282 LOC: Adaptation::Icap::TheConfig.io_timeout_raw
9285 This parameter specifies how long to wait for an I/O activity on
9286 an established, active ICAP connection before giving up and
9287 either terminating the HTTP transaction or bypassing the
9291 NAME: icap_service_failure_limit
9292 COMMENT: limit [in memory-depth time-units]
9293 TYPE: icap_service_failure_limit
9295 LOC: Adaptation::Icap::TheConfig
9298 The limit specifies the number of failures that Squid tolerates
9299 when establishing a new TCP connection with an ICAP service. If
9300 the number of failures exceeds the limit, the ICAP service is
9301 not used for new ICAP requests until it is time to refresh its
9304 A negative value disables the limit. Without the limit, an ICAP
9305 service will not be considered down due to connectivity failures
9306 between ICAP OPTIONS requests.
9308 Squid forgets ICAP service failures older than the specified
9309 value of memory-depth. The memory fading algorithm
9310 is approximate because Squid does not remember individual
9311 errors but groups them instead, splitting the option
9312 value into ten time slots of equal length.
9314 When memory-depth is 0 and by default this option has no
9315 effect on service failure expiration.
9317 Squid always forgets failures when updating service settings
9318 using an ICAP OPTIONS transaction, regardless of this option
9322 # suspend service usage after 10 failures in 5 seconds:
9323 icap_service_failure_limit 10 in 5 seconds
9326 NAME: icap_service_revival_delay
9329 LOC: Adaptation::Icap::TheConfig.service_revival_delay
9332 The delay specifies the number of seconds to wait after an ICAP
9333 OPTIONS request failure before requesting the options again. The
9334 failed ICAP service is considered "down" until fresh OPTIONS are
9337 The actual delay cannot be smaller than the hardcoded minimum
9338 delay of 30 seconds.
9341 NAME: icap_preview_enable
9345 LOC: Adaptation::Icap::TheConfig.preview_enable
9348 The ICAP Preview feature allows the ICAP server to handle the
9349 HTTP message by looking only at the beginning of the message body
9350 or even without receiving the body at all. In some environments,
9351 previews greatly speedup ICAP processing.
9353 During an ICAP OPTIONS transaction, the server may tell Squid what
9354 HTTP messages should be previewed and how big the preview should be.
9355 Squid will not use Preview if the server did not request one.
9357 To disable ICAP Preview for all ICAP services, regardless of
9358 individual ICAP server OPTIONS responses, set this option to "off".
9360 icap_preview_enable off
9363 NAME: icap_preview_size
9366 LOC: Adaptation::Icap::TheConfig.preview_size
9368 DEFAULT_DOC: No preview sent.
9370 The default size of preview data to be sent to the ICAP server.
9371 This value might be overwritten on a per server basis by OPTIONS requests.
9374 NAME: icap_206_enable
9378 LOC: Adaptation::Icap::TheConfig.allow206_enable
9381 206 (Partial Content) responses is an ICAP extension that allows the
9382 ICAP agents to optionally combine adapted and original HTTP message
9383 content. The decision to combine is postponed until the end of the
9384 ICAP response. Squid supports Partial Content extension by default.
9386 Activation of the Partial Content extension is negotiated with each
9387 ICAP service during OPTIONS exchange. Most ICAP servers should handle
9388 negotiation correctly even if they do not support the extension, but
9389 some might fail. To disable Partial Content support for all ICAP
9390 services and to avoid any negotiation, set this option to "off".
9396 NAME: icap_default_options_ttl
9399 LOC: Adaptation::Icap::TheConfig.default_options_ttl
9402 The default TTL value for ICAP OPTIONS responses that don't have
9403 an Options-TTL header.
9406 NAME: icap_persistent_connections
9410 LOC: Adaptation::Icap::TheConfig.reuse_connections
9413 Whether or not Squid should use persistent connections to
9417 NAME: adaptation_send_client_ip icap_send_client_ip
9419 IFDEF: USE_ADAPTATION
9421 LOC: Adaptation::Config::send_client_ip
9424 If enabled, Squid shares HTTP client IP information with adaptation
9425 services. For ICAP, Squid adds the X-Client-IP header to ICAP requests.
9426 For eCAP, Squid sets the libecap::metaClientIp transaction option.
9428 See also: adaptation_uses_indirect_client
9431 NAME: adaptation_send_username icap_send_client_username
9433 IFDEF: USE_ADAPTATION
9435 LOC: Adaptation::Config::send_username
9438 This sends authenticated HTTP client username (if available) to
9439 the adaptation service.
9441 For ICAP, the username value is encoded based on the
9442 icap_client_username_encode option and is sent using the header
9443 specified by the icap_client_username_header option.
9446 NAME: icap_client_username_header
9449 LOC: Adaptation::Icap::TheConfig.client_username_header
9450 DEFAULT: X-Client-Username
9452 ICAP request header name to use for adaptation_send_username.
9455 NAME: icap_client_username_encode
9459 LOC: Adaptation::Icap::TheConfig.client_username_encode
9462 Whether to base64 encode the authenticated client username.
9466 TYPE: icap_service_type
9468 LOC: Adaptation::Icap::TheConfig
9471 Defines a single ICAP service using the following format:
9473 icap_service id vectoring_point uri [option ...]
9476 an opaque identifier or name which is used to direct traffic to
9477 this specific service. Must be unique among all adaptation
9478 services in squid.conf.
9480 vectoring_point: reqmod_precache|reqmod_postcache|respmod_precache|respmod_postcache
9481 This specifies at which point of transaction processing the
9482 ICAP service should be activated. *_postcache vectoring points
9483 are not yet supported.
9485 uri: icap://servername:port/servicepath
9486 ICAP server and service location.
9487 icaps://servername:port/servicepath
9488 The "icap:" URI scheme is used for traditional ICAP server and
9489 service location (default port is 1344, connections are not
9490 encrypted). The "icaps:" URI scheme is for Secure ICAP
9491 services that use SSL/TLS-encrypted ICAP connections (by
9492 default, on port 11344).
9494 ICAP does not allow a single service to handle both REQMOD and RESPMOD
9495 transactions. Squid does not enforce that requirement. You can specify
9496 services with the same service_url and different vectoring_points. You
9497 can even specify multiple identical services as long as their
9498 service_names differ.
9500 To activate a service, use the adaptation_access directive. To group
9501 services, use adaptation_service_chain and adaptation_service_set.
9503 Service options are separated by white space. ICAP services support
9504 the following name=value options:
9507 If set to 'on' or '1', the ICAP service is treated as
9508 optional. If the service cannot be reached or malfunctions,
9509 Squid will try to ignore any errors and process the message as
9510 if the service was not enabled. No all ICAP errors can be
9511 bypassed. If set to 0, the ICAP service is treated as
9512 essential and all ICAP errors will result in an error page
9513 returned to the HTTP client.
9515 Bypass is off by default: services are treated as essential.
9518 If set to 'on' or '1', the ICAP service is allowed to
9519 dynamically change the current message adaptation plan by
9520 returning a chain of services to be used next. The services
9521 are specified using the X-Next-Services ICAP response header
9522 value, formatted as a comma-separated list of service names.
9523 Each named service should be configured in squid.conf. Other
9524 services are ignored. An empty X-Next-Services value results
9525 in an empty plan which ends the current adaptation.
9527 Dynamic adaptation plan may cross or cover multiple supported
9528 vectoring points in their natural processing order.
9530 Routing is not allowed by default: the ICAP X-Next-Services
9531 response header is ignored.
9534 Only has effect on split-stack systems. The default on those systems
9535 is to use IPv4-only connections. When set to 'on' this option will
9536 make Squid use IPv6-only connections to contact this ICAP service.
9538 on-overload=block|bypass|wait|force
9539 If the service Max-Connections limit has been reached, do
9540 one of the following for each new ICAP transaction:
9541 * block: send an HTTP error response to the client
9542 * bypass: ignore the "over-connected" ICAP service
9543 * wait: wait (in a FIFO queue) for an ICAP connection slot
9544 * force: proceed, ignoring the Max-Connections limit
9546 In SMP mode with N workers, each worker assumes the service
9547 connection limit is Max-Connections/N, even though not all
9548 workers may use a given service.
9550 The default value is "bypass" if service is bypassable,
9551 otherwise it is set to "wait".
9555 Use the given number as the Max-Connections limit, regardless
9556 of the Max-Connections value given by the service, if any.
9558 connection-encryption=on|off
9559 Determines the ICAP service effect on the connections_encrypted
9562 The default is "on" for Secure ICAP services (i.e., those
9563 with the icaps:// service URIs scheme) and "off" for plain ICAP
9566 Does not affect ICAP connections (e.g., does not turn Secure
9569 ==== ICAPS / TLS OPTIONS ====
9571 These options are used for Secure ICAP (icaps://....) services only.
9573 tls-cert=/path/to/ssl/certificate
9574 A client X.509 certificate to use when connecting to
9577 tls-key=/path/to/ssl/key
9578 The private key corresponding to the previous
9581 If tls-key= is not specified tls-cert= is assumed to
9582 reference a PEM file containing both the certificate
9585 tls-cipher=... The list of valid TLS/SSL ciphers to use when connecting
9586 to this icap server.
9589 The minimum TLS protocol version to permit. To control
9590 SSLv3 use the tls-options= parameter.
9591 Supported Values: 1.0 (default), 1.1, 1.2
9593 tls-options=... Specify various OpenSSL library options:
9595 NO_SSLv3 Disallow the use of SSLv3
9598 Always create a new key when using
9599 temporary/ephemeral DH key exchanges
9601 ALL Enable various bug workarounds
9602 suggested as "harmless" by OpenSSL
9603 Be warned that this reduces SSL/TLS
9604 strength to some attacks.
9606 See the OpenSSL SSL_CTX_set_options documentation for a
9607 more complete list. Options relevant only to SSLv2 are
9610 tls-cafile= PEM file containing CA certificates to use when verifying
9611 the icap server certificate.
9612 Use to specify intermediate CA certificate(s) if not sent
9613 by the server. Or the full CA chain for the server when
9614 using the tls-default-ca=off flag.
9615 May be repeated to load multiple files.
9617 tls-capath=... A directory containing additional CA certificates to
9618 use when verifying the icap server certificate.
9619 Requires OpenSSL or LibreSSL.
9621 tls-crlfile=... A certificate revocation list file to use when
9622 verifying the icap server certificate.
9624 tls-flags=... Specify various flags modifying the Squid TLS implementation:
9627 Accept certificates even if they fail to
9630 Don't verify the icap server certificate
9631 matches the server name
9633 tls-default-ca[=off]
9634 Whether to use the system Trusted CAs. Default is ON.
9636 tls-domain= The icap server name as advertised in it's certificate.
9637 Used for verifying the correctness of the received icap
9638 server certificate. If not specified the icap server
9639 hostname extracted from ICAP URI will be used.
9641 Older icap_service format without optional named parameters is
9642 deprecated but supported for backward compatibility.
9645 icap_service svcBlocker reqmod_precache icap://icap1.mydomain.net:1344/reqmod bypass=0
9646 icap_service svcLogger reqmod_precache icaps://icap2.mydomain.net:11344/reqmod routing=on
9650 TYPE: icap_class_type
9655 This deprecated option was documented to define an ICAP service
9656 chain, even though it actually defined a set of similar, redundant
9657 services, and the chains were not supported.
9659 To define a set of redundant services, please use the
9660 adaptation_service_set directive. For service chains, use
9661 adaptation_service_chain.
9665 TYPE: icap_access_type
9670 This option is deprecated. Please use adaptation_access, which
9671 has the same ICAP functionality, but comes with better
9672 documentation, and eCAP support.
9677 -----------------------------------------------------------------------------
9684 LOC: Adaptation::Ecap::TheConfig.onoff
9687 Controls whether eCAP support is enabled.
9691 TYPE: ecap_service_type
9693 LOC: Adaptation::Ecap::TheConfig
9696 Defines a single eCAP service
9698 ecap_service id vectoring_point uri [option ...]
9701 an opaque identifier or name which is used to direct traffic to
9702 this specific service. Must be unique among all adaptation
9703 services in squid.conf.
9705 vectoring_point: reqmod_precache|reqmod_postcache|respmod_precache|respmod_postcache
9706 This specifies at which point of transaction processing the
9707 eCAP service should be activated. *_postcache vectoring points
9708 are not yet supported.
9710 uri: ecap://vendor/service_name?custom&cgi=style¶meters=optional
9711 Squid uses the eCAP service URI to match this configuration
9712 line with one of the dynamically loaded services. Each loaded
9713 eCAP service must have a unique URI. Obtain the right URI from
9714 the service provider.
9716 To activate a service, use the adaptation_access directive. To group
9717 services, use adaptation_service_chain and adaptation_service_set.
9719 Service options are separated by white space. eCAP services support
9720 the following name=value options:
9723 If set to 'on' or '1', the eCAP service is treated as optional.
9724 If the service cannot be reached or malfunctions, Squid will try
9725 to ignore any errors and process the message as if the service
9726 was not enabled. No all eCAP errors can be bypassed.
9727 If set to 'off' or '0', the eCAP service is treated as essential
9728 and all eCAP errors will result in an error page returned to the
9731 Bypass is off by default: services are treated as essential.
9734 If set to 'on' or '1', the eCAP service is allowed to
9735 dynamically change the current message adaptation plan by
9736 returning a chain of services to be used next.
9738 Dynamic adaptation plan may cross or cover multiple supported
9739 vectoring points in their natural processing order.
9741 Routing is not allowed by default.
9743 connection-encryption=on|off
9744 Determines the eCAP service effect on the connections_encrypted
9747 Defaults to "on", which does not taint the master transaction
9750 Does not affect eCAP API calls.
9752 Older ecap_service format without optional named parameters is
9753 deprecated but supported for backward compatibility.
9757 ecap_service s1 reqmod_precache ecap://filters.R.us/leakDetector?on_error=block bypass=off
9758 ecap_service s2 respmod_precache ecap://filters.R.us/virusFilter config=/etc/vf.cfg bypass=on
9761 NAME: loadable_modules
9763 IFDEF: USE_LOADABLE_MODULES
9764 LOC: Config.loadable_module_names
9767 Instructs Squid to load the specified dynamic module(s) or activate
9768 preloaded module(s).
9770 loadable_modules @DEFAULT_PREFIX@/lib/MinimalAdapter.so
9774 MESSAGE ADAPTATION OPTIONS
9775 -----------------------------------------------------------------------------
9778 NAME: adaptation_service_set
9779 TYPE: adaptation_service_set_type
9780 IFDEF: USE_ADAPTATION
9785 Configures an ordered set of similar, redundant services. This is
9786 useful when hot standby or backup adaptation servers are available.
9788 adaptation_service_set set_name service_name1 service_name2 ...
9790 The named services are used in the set declaration order. The first
9791 applicable adaptation service from the set is used first. The next
9792 applicable service is tried if and only if the transaction with the
9793 previous service fails and the message waiting to be adapted is still
9796 When adaptation starts, broken services are ignored as if they were
9797 not a part of the set. A broken service is a down optional service.
9799 The services in a set must be attached to the same vectoring point
9800 (e.g., pre-cache) and use the same adaptation method (e.g., REQMOD).
9802 If all services in a set are optional then adaptation failures are
9803 bypassable. If all services in the set are essential, then a
9804 transaction failure with one service may still be retried using
9805 another service from the set, but when all services fail, the master
9806 transaction fails as well.
9808 A set may contain a mix of optional and essential services, but that
9809 is likely to lead to surprising results because broken services become
9810 ignored (see above), making previously bypassable failures fatal.
9811 Technically, it is the bypassability of the last failed service that
9814 See also: adaptation_access adaptation_service_chain
9817 adaptation_service_set svcBlocker urlFilterPrimary urlFilterBackup
9818 adaptation service_set svcLogger loggerLocal loggerRemote
9821 NAME: adaptation_service_chain
9822 TYPE: adaptation_service_chain_type
9823 IFDEF: USE_ADAPTATION
9828 Configures a list of complementary services that will be applied
9829 one-by-one, forming an adaptation chain or pipeline. This is useful
9830 when Squid must perform different adaptations on the same message.
9832 adaptation_service_chain chain_name service_name1 svc_name2 ...
9834 The named services are used in the chain declaration order. The first
9835 applicable adaptation service from the chain is used first. The next
9836 applicable service is applied to the successful adaptation results of
9837 the previous service in the chain.
9839 When adaptation starts, broken services are ignored as if they were
9840 not a part of the chain. A broken service is a down optional service.
9842 Request satisfaction terminates the adaptation chain because Squid
9843 does not currently allow declaration of RESPMOD services at the
9844 "reqmod_precache" vectoring point (see icap_service or ecap_service).
9846 The services in a chain must be attached to the same vectoring point
9847 (e.g., pre-cache) and use the same adaptation method (e.g., REQMOD).
9849 A chain may contain a mix of optional and essential services. If an
9850 essential adaptation fails (or the failure cannot be bypassed for
9851 other reasons), the master transaction fails. Otherwise, the failure
9852 is bypassed as if the failed adaptation service was not in the chain.
9854 See also: adaptation_access adaptation_service_set
9857 adaptation_service_chain svcRequest requestLogger urlFilter leakDetector
9860 NAME: adaptation_access
9861 TYPE: adaptation_access_type
9862 IFDEF: USE_ADAPTATION
9865 DEFAULT_DOC: Allow, unless rules exist in squid.conf.
9867 Sends an HTTP transaction to an ICAP or eCAP adaptation service.
9869 adaptation_access service_name allow|deny [!]aclname...
9870 adaptation_access set_name allow|deny [!]aclname...
9872 At each supported vectoring point, the adaptation_access
9873 statements are processed in the order they appear in this
9874 configuration file. Statements pointing to the following services
9875 are ignored (i.e., skipped without checking their ACL):
9877 - services serving different vectoring points
9878 - "broken-but-bypassable" services
9879 - "up" services configured to ignore such transactions
9880 (e.g., based on the ICAP Transfer-Ignore header).
9882 When a set_name is used, all services in the set are checked
9883 using the same rules, to find the first applicable one. See
9884 adaptation_service_set for details.
9886 If an access list is checked and there is a match, the
9887 processing stops: For an "allow" rule, the corresponding
9888 adaptation service is used for the transaction. For a "deny"
9889 rule, no adaptation service is activated.
9891 It is currently not possible to apply more than one adaptation
9892 service at the same vectoring point to the same HTTP transaction.
9894 See also: icap_service and ecap_service
9897 adaptation_access service_1 allow all
9900 NAME: adaptation_service_iteration_limit
9902 IFDEF: USE_ADAPTATION
9903 LOC: Adaptation::Config::service_iteration_limit
9906 Limits the number of iterations allowed when applying adaptation
9907 services to a message. If your longest adaptation set or chain
9908 may have more than 16 services, increase the limit beyond its
9909 default value of 16. If detecting infinite iteration loops sooner
9910 is critical, make the iteration limit match the actual number
9911 of services in your longest adaptation set or chain.
9913 Infinite adaptation loops are most likely with routing services.
9915 See also: icap_service routing=1
9918 NAME: adaptation_masterx_shared_names
9920 IFDEF: USE_ADAPTATION
9921 LOC: Adaptation::Config::masterx_shared_name
9924 For each master transaction (i.e., the HTTP request and response
9925 sequence, including all related ICAP and eCAP exchanges), Squid
9926 maintains a table of metadata. The table entries are (name, value)
9927 pairs shared among eCAP and ICAP exchanges. The table is destroyed
9928 with the master transaction.
9930 This option specifies the table entry names that Squid must accept
9931 from and forward to the adaptation transactions.
9933 An ICAP REQMOD or RESPMOD transaction may set an entry in the
9934 shared table by returning an ICAP header field with a name
9935 specified in adaptation_masterx_shared_names.
9937 An eCAP REQMOD or RESPMOD transaction may set an entry in the
9938 shared table by implementing the libecap::visitEachOption() API
9939 to provide an option with a name specified in
9940 adaptation_masterx_shared_names.
9942 Squid will store and forward the set entry to subsequent adaptation
9943 transactions within the same master transaction scope.
9945 Only one shared entry name is supported at this time.
9948 # share authentication information among ICAP services
9949 adaptation_masterx_shared_names X-Subscriber-ID
9952 NAME: adaptation_meta
9954 IFDEF: USE_ADAPTATION
9955 LOC: Adaptation::Config::metaHeaders()
9958 This option allows Squid administrator to add custom ICAP request
9959 headers or eCAP options to Squid ICAP requests or eCAP transactions.
9960 Use it to pass custom authentication tokens and other
9961 transaction-state related meta information to an ICAP/eCAP service.
9963 The addition of a meta header is ACL-driven:
9964 adaptation_meta name value [!]aclname ...
9966 Processing for a given header name stops after the first ACL list match.
9967 Thus, it is impossible to add two headers with the same name. If no ACL
9968 lists match for a given header name, no such header is added. For
9971 # do not debug transactions except for those that need debugging
9972 adaptation_meta X-Debug 1 needs_debugging
9974 # log all transactions except for those that must remain secret
9975 adaptation_meta X-Log 1 !keep_secret
9977 # mark transactions from users in the "G 1" group
9978 adaptation_meta X-Authenticated-Groups "G 1" authed_as_G1
9980 The "value" parameter may be a regular squid.conf token or a "double
9981 quoted string". Within the quoted string, use backslash (\) to escape
9982 any character, which is currently only useful for escaping backslashes
9983 and double quotes. For example,
9984 "this string has one backslash (\\) and two \"quotes\""
9986 Used adaptation_meta header values may be logged via %note
9987 logformat code. If multiple adaptation_meta headers with the same name
9988 are used during master transaction lifetime, the header values are
9989 logged in the order they were used and duplicate values are ignored
9990 (only the first repeated value will be logged).
9996 LOC: Adaptation::Icap::TheConfig.repeat
9997 DEFAULT_IF_NONE: deny all
9999 This ACL determines which retriable ICAP transactions are
10000 retried. Transactions that received a complete ICAP response
10001 and did not have to consume or produce HTTP bodies to receive
10002 that response are usually retriable.
10004 icap_retry allow|deny [!]aclname ...
10006 Squid automatically retries some ICAP I/O timeouts and errors
10007 due to persistent connection race conditions.
10009 See also: icap_retry_limit
10012 NAME: icap_retry_limit
10015 LOC: Adaptation::Icap::TheConfig.repeat_limit
10017 DEFAULT_DOC: No retries are allowed.
10019 Limits the number of retries allowed.
10021 Communication errors due to persistent connection race
10022 conditions are unavoidable, automatically retried, and do not
10023 count against this limit.
10025 See also: icap_retry
10031 -----------------------------------------------------------------------------
10034 NAME: check_hostnames
10037 LOC: Config.onoff.check_hostnames
10039 For security and stability reasons Squid can check
10040 hostnames for Internet standard RFC compliance. If you want
10041 Squid to perform these checks turn this directive on.
10044 NAME: allow_underscore
10047 LOC: Config.onoff.allow_underscore
10049 Underscore characters is not strictly allowed in Internet hostnames
10050 but nevertheless used by many sites. Set this to off if you want
10051 Squid to be strict about the standard.
10052 This check is performed only when check_hostnames is set to on.
10055 NAME: dns_retransmit_interval
10058 LOC: Config.Timeout.idns_retransmit
10060 Initial retransmit interval for DNS queries. The interval is
10061 doubled each time all configured DNS servers have been tried.
10066 DEFAULT: 30 seconds
10067 LOC: Config.Timeout.idns_query
10069 DNS Query timeout. If no response is received to a DNS query
10070 within this time all DNS servers for the queried domain
10071 are assumed to be unavailable.
10074 NAME: dns_packet_max
10076 DEFAULT_DOC: EDNS disabled
10078 LOC: Config.dns.packet_max
10080 Maximum number of bytes packet size to advertise via EDNS.
10081 Set to "none" to disable EDNS large packet support.
10083 For legacy reasons DNS UDP replies will default to 512 bytes which
10084 is too small for many responses. EDNS provides a means for Squid to
10085 negotiate receiving larger responses back immediately without having
10086 to failover with repeat requests. Responses larger than this limit
10087 will retain the old behaviour of failover to TCP DNS.
10089 Squid has no real fixed limit internally, but allowing packet sizes
10090 over 1500 bytes requires network jumbogram support and is usually not
10093 WARNING: The RFC also indicates that some older resolvers will reply
10094 with failure of the whole request if the extension is added. Some
10095 resolvers have already been identified which will reply with mangled
10096 EDNS response on occasion. Usually in response to many-KB jumbogram
10097 sizes being advertised by Squid.
10098 Squid will currently treat these both as an unable-to-resolve domain
10099 even if it would be resolvable without EDNS.
10106 DEFAULT_DOC: Search for single-label domain names is disabled.
10107 LOC: Config.onoff.res_defnames
10109 Normally the RES_DEFNAMES resolver option is disabled
10110 (see res_init(3)). This prevents caches in a hierarchy
10111 from interpreting single-component hostnames locally. To allow
10112 Squid to handle single-component names, enable this option.
10115 NAME: dns_multicast_local
10119 DEFAULT_DOC: Search for .local and .arpa names is disabled.
10120 LOC: Config.onoff.dns_mdns
10122 When set to on, Squid sends multicast DNS lookups on the local
10123 network for domains ending in .local and .arpa.
10124 This enables local servers and devices to be contacted in an
10125 ad-hoc or zero-configuration network environment.
10128 NAME: dns_nameservers
10131 DEFAULT_DOC: Use operating system definitions
10132 LOC: Config.dns.nameservers
10134 Use this if you want to specify a list of DNS name servers
10135 (IP addresses) to use instead of those given in your
10136 /etc/resolv.conf file.
10138 On Windows platforms, if no value is specified here or in
10139 the /etc/resolv.conf file, the list of DNS name servers are
10140 taken from the Windows registry, both static and dynamic DHCP
10141 configurations are supported.
10143 Example: dns_nameservers 10.0.0.1 192.172.0.4
10148 DEFAULT: @DEFAULT_HOSTS@
10149 LOC: Config.etcHostsPath
10151 Location of the host-local IP name-address associations
10152 database. Most Operating Systems have such a file on different
10154 - Un*X & Linux: /etc/hosts
10155 - Windows NT/2000: %SystemRoot%\system32\drivers\etc\hosts
10156 (%SystemRoot% value install default is c:\winnt)
10157 - Windows XP/2003: %SystemRoot%\system32\drivers\etc\hosts
10158 (%SystemRoot% value install default is c:\windows)
10159 - Windows 9x/Me: %windir%\hosts
10160 (%windir% value is usually c:\windows)
10161 - Cygwin: /etc/hosts
10163 The file contains newline-separated definitions, in the
10164 form ip_address_in_dotted_form name [name ...] names are
10165 whitespace-separated. Lines beginning with an hash (#)
10166 character are comments.
10168 The file is checked at startup and upon configuration.
10169 If set to 'none', it won't be checked.
10170 If append_domain is used, that domain will be added to
10171 domain-local (i.e. not containing any dot character) host
10175 NAME: append_domain
10177 LOC: Config.appendDomain
10179 DEFAULT_DOC: Use operating system definitions
10181 Appends local domain name to hostnames without any dots in
10182 them. append_domain must begin with a period.
10184 Be warned there are now Internet names with no dots in
10185 them using only top-domain names, so setting this may
10186 cause some Internet sites to become unavailable.
10189 append_domain .yourdomain.com
10192 NAME: ignore_unknown_nameservers
10194 LOC: Config.onoff.ignore_unknown_nameservers
10197 By default Squid checks that DNS responses are received
10198 from the same IP addresses they are sent to. If they
10199 don't match, Squid ignores the response and writes a warning
10200 message to cache.log. You can allow responses from unknown
10201 nameservers by setting this option to 'off'.
10205 COMMENT: (number of entries)
10208 LOC: Config.ipcache.size
10210 Maximum number of DNS IP cache entries.
10217 LOC: Config.ipcache.low
10224 LOC: Config.ipcache.high
10226 The size, low-, and high-water marks for the IP cache.
10229 NAME: fqdncache_size
10230 COMMENT: (number of entries)
10233 LOC: Config.fqdncache.size
10235 Maximum number of FQDN cache entries.
10240 -----------------------------------------------------------------------------
10243 NAME: configuration_includes_quoted_values
10245 TYPE: configuration_includes_quoted_values
10247 LOC: ConfigParser::RecognizeQuotedValues
10249 If set, Squid will recognize each "quoted string" after a configuration
10250 directive as a single parameter. The quotes are stripped before the
10251 parameter value is interpreted or used.
10252 See "Values with spaces, quotes, and other special characters"
10253 section for more details.
10260 LOC: Config.onoff.mem_pools
10262 If set, Squid will keep pools of allocated (but unused) memory
10263 available for future use. If memory is a premium on your
10264 system and you believe your malloc library outperforms Squid
10265 routines, disable this.
10268 NAME: memory_pools_limit
10272 LOC: Config.MemPools.limit
10274 Used only with memory_pools on:
10275 memory_pools_limit 50 MB
10277 If set to a non-zero value, Squid will keep at most the specified
10278 limit of allocated (but unused) memory in memory pools. All free()
10279 requests that exceed this limit will be handled by your malloc
10280 library. Squid does not pre-allocate any memory, just safe-keeps
10281 objects that otherwise would be free()d. Thus, it is safe to set
10282 memory_pools_limit to a reasonably high value even if your
10283 configuration will use less memory.
10285 If set to none, Squid will keep all memory it can. That is, there
10286 will be no limit on the total amount of memory used for safe-keeping.
10288 To disable memory allocation optimization, do not set
10289 memory_pools_limit to 0 or none. Set memory_pools to "off" instead.
10291 An overhead for maintaining memory pools is not taken into account
10292 when the limit is checked. This overhead is close to four bytes per
10293 object kept. However, pools may actually _save_ memory because of
10294 reduced memory thrashing in your malloc library.
10297 NAME: forwarded_for
10298 COMMENT: on|off|transparent|truncate|delete
10301 LOC: opt_forwarded_for
10303 If set to "on", Squid will append your client's IP address
10304 in the HTTP requests it forwards. By default it looks like:
10306 X-Forwarded-For: 192.1.2.3
10308 If set to "off", it will appear as
10310 X-Forwarded-For: unknown
10312 If set to "transparent", Squid will not alter the
10313 X-Forwarded-For header in any way.
10315 If set to "delete", Squid will delete the entire
10316 X-Forwarded-For header.
10318 If set to "truncate", Squid will remove all existing
10319 X-Forwarded-For entries, and place the client IP as the sole entry.
10322 NAME: cachemgr_passwd
10323 TYPE: cachemgrpasswd
10325 DEFAULT_DOC: No password. Actions which require password are denied.
10326 LOC: Config.passwd_list
10328 Specify passwords for cachemgr operations.
10330 Usage: cachemgr_passwd password action action ...
10332 Some valid actions are (see cache manager menu for a full list):
10370 * Indicates actions which will not be performed without a
10371 valid password, others can be performed if not listed here.
10373 To disable an action, set the password to "disable".
10374 To allow performing an action without a password, set the
10375 password to "none".
10377 Use the keyword "all" to set the same password for all actions.
10380 cachemgr_passwd secret shutdown
10381 cachemgr_passwd lesssssssecret info stats/objects
10382 cachemgr_passwd disable all
10389 LOC: Config.onoff.client_db
10391 If you want to disable collecting per-client statistics,
10392 turn off client_db here.
10395 NAME: refresh_all_ims
10399 LOC: Config.onoff.refresh_all_ims
10401 When you enable this option, squid will always check
10402 the origin server for an update when a client sends an
10403 If-Modified-Since request. Many browsers use IMS
10404 requests when the user requests a reload, and this
10405 ensures those clients receive the latest version.
10407 By default (off), squid may return a Not Modified response
10408 based on the age of the cached version.
10411 NAME: reload_into_ims
10412 IFDEF: USE_HTTP_VIOLATIONS
10416 LOC: Config.onoff.reload_into_ims
10418 When you enable this option, client no-cache or ``reload''
10419 requests will be changed to If-Modified-Since requests.
10420 Doing this VIOLATES the HTTP standard. Enabling this
10421 feature could make you liable for problems which it
10424 see also refresh_pattern for a more selective approach.
10427 NAME: connect_retries
10429 LOC: Config.connect_retries
10431 DEFAULT_DOC: Do not retry failed connections.
10433 Limits the number of reopening attempts when establishing a single
10434 TCP connection. All these attempts must still complete before the
10435 applicable connection opening timeout expires.
10437 By default and when connect_retries is set to zero, Squid does not
10438 retry failed connection opening attempts.
10440 The (not recommended) maximum is 10 tries. An attempt to configure a
10441 higher value results in the value of 10 being used (with a warning).
10443 Squid may open connections to retry various high-level forwarding
10444 failures. For an outside observer, that activity may look like a
10445 low-level connection reopening attempt, but those high-level retries
10446 are governed by forward_max_tries instead.
10448 See also: connect_timeout, forward_timeout, icap_connect_timeout,
10449 and forward_max_tries.
10452 NAME: retry_on_error
10454 LOC: Config.retry.onerror
10457 If set to ON Squid will automatically retry requests when
10458 receiving an error response with status 403 (Forbidden),
10459 500 (Internal Error), 501 or 503 (Service not available).
10460 Status 502 and 504 (Gateway errors) are always retried.
10462 This is mainly useful if you are in a complex cache hierarchy to
10463 work around access control errors.
10465 NOTE: This retry will attempt to find another working destination.
10466 Which is different from the server which just failed.
10469 NAME: as_whois_server
10471 LOC: Config.as_whois_server
10472 DEFAULT: whois.ra.net
10474 WHOIS server to query for AS numbers. NOTE: AS numbers are
10475 queried only when Squid starts up, not for every request.
10480 LOC: Config.onoff.offline
10483 Enable this option and Squid will never try to validate cached
10487 NAME: uri_whitespace
10488 TYPE: uri_whitespace
10489 LOC: Config.uri_whitespace
10492 What to do with requests that have whitespace characters in the
10495 strip: The whitespace characters are stripped out of the URL.
10496 This is the behavior recommended by RFC2396 and RFC3986
10497 for tolerant handling of generic URI.
10498 NOTE: This is one difference between generic URI and HTTP URLs.
10500 deny: The request is denied. The user receives an "Invalid
10502 This is the behaviour recommended by RFC2616 for safe
10503 handling of HTTP request URL.
10505 allow: The request is allowed and the URI is not changed. The
10506 whitespace characters remain in the URI. Note the
10507 whitespace is passed to redirector processes if they
10509 Note this may be considered a violation of RFC2616
10510 request parsing where whitespace is prohibited in the
10513 encode: The request is allowed and the whitespace characters are
10514 encoded according to RFC1738.
10516 chop: The request is allowed and the URI is chopped at the
10520 NOTE the current Squid implementation of encode and chop violates
10521 RFC2616 by not using a 301 redirect after altering the URL.
10526 LOC: Config.chroot_dir
10529 Specifies a directory where Squid should do a chroot() while
10530 initializing. This also causes Squid to fully drop root
10531 privileges after initializing. This means, for example, if you
10532 use a HTTP port less than 1024 and try to reconfigure, you may
10533 get an error saying that Squid can not open the port.
10536 NAME: pipeline_prefetch
10537 TYPE: pipelinePrefetch
10538 LOC: Config.pipeline_max_prefetch
10540 DEFAULT_DOC: Do not pre-parse pipelined requests.
10542 HTTP clients may send a pipeline of 1+N requests to Squid using a
10543 single connection, without waiting for Squid to respond to the first
10544 of those requests. This option limits the number of concurrent
10545 requests Squid will try to handle in parallel. If set to N, Squid
10546 will try to receive and process up to 1+N requests on the same
10547 connection concurrently.
10549 Defaults to 0 (off) for bandwidth management and access logging
10552 NOTE: pipelining requires persistent connections to clients.
10554 WARNING: pipelining breaks NTLM and Negotiate/Kerberos authentication.
10557 NAME: high_response_time_warning
10560 LOC: Config.warnings.high_rptm
10562 DEFAULT_DOC: disabled.
10564 If the one-minute median response time exceeds this value,
10565 Squid prints a WARNING with debug level 0 to get the
10566 administrators attention. The value is in milliseconds.
10569 NAME: high_page_fault_warning
10571 LOC: Config.warnings.high_pf
10573 DEFAULT_DOC: disabled.
10575 If the one-minute average page fault rate exceeds this
10576 value, Squid prints a WARNING with debug level 0 to get
10577 the administrators attention. The value is in page faults
10581 NAME: high_memory_warning
10583 LOC: Config.warnings.high_memory
10584 IFDEF: HAVE_MSTATS&&HAVE_GNUMALLOC_H
10586 DEFAULT_DOC: disabled.
10588 If the memory usage (as determined by gnumalloc, if available and used)
10589 exceeds this amount, Squid prints a WARNING with debug level 0 to get
10590 the administrators attention.
10592 # TODO: link high_memory_warning to mempools?
10594 NAME: sleep_after_fork
10595 COMMENT: (microseconds)
10597 LOC: Config.sleep_after_fork
10600 When this is set to a non-zero value, the main Squid process
10601 sleeps the specified number of microseconds after a fork()
10602 system call. This sleep may help the situation where your
10603 system reports fork() failures due to lack of (virtual)
10604 memory. Note, however, if you have a lot of child
10605 processes, these sleep delays will add up and your
10606 Squid will not service requests for some amount of time
10607 until all the child processes have been started.
10610 NAME: windows_ipaddrchangemonitor
10611 IFDEF: _SQUID_WINDOWS_
10615 LOC: Config.onoff.WIN32_IpAddrChangeMonitor
10617 On Windows Squid by default will monitor IP address changes and will
10618 reconfigure itself after any detected event. This is very useful for
10619 proxies connected to internet with dial-up interfaces.
10620 In some cases (a Proxy server acting as VPN gateway is one) it could be
10621 desiderable to disable this behaviour setting this to 'off'.
10622 Note: after changing this, Squid service must be restarted.
10627 IFDEF: USE_SQUID_EUI
10629 LOC: Eui::TheConfig.euiLookup
10631 Whether to lookup the EUI or MAC address of a connected client.
10634 NAME: max_filedescriptors max_filedesc
10637 DEFAULT_DOC: Use operating system soft limit set by ulimit.
10638 LOC: Config.max_filedescriptors
10640 Set the maximum number of filedescriptors, either below the
10641 operating system default or up to the hard limit.
10643 Remove from squid.conf to inherit the current ulimit soft
10646 Note: Changing this requires a restart of Squid. Also
10647 not all I/O types supports large values (eg on Windows).
10650 NAME: force_request_body_continuation
10652 LOC: Config.accessList.forceRequestBodyContinuation
10654 DEFAULT_DOC: Deny, unless rules exist in squid.conf.
10656 This option controls how Squid handles data upload requests from HTTP
10657 and FTP agents that require a "Please Continue" control message response
10658 to actually send the request body to Squid. It is mostly useful in
10659 adaptation environments.
10661 When Squid receives an HTTP request with an "Expect: 100-continue"
10662 header or an FTP upload command (e.g., STOR), Squid normally sends the
10663 request headers or FTP command information to an adaptation service (or
10664 peer) and waits for a response. Most adaptation services (and some
10665 broken peers) may not respond to Squid at that stage because they may
10666 decide to wait for the HTTP request body or FTP data transfer. However,
10667 that request body or data transfer may never come because Squid has not
10668 responded with the HTTP 100 or FTP 150 (Please Continue) control message
10669 to the request sender yet!
10671 An allow match tells Squid to respond with the HTTP 100 or FTP 150
10672 (Please Continue) control message on its own, before forwarding the
10673 request to an adaptation service or peer. Such a response usually forces
10674 the request sender to proceed with sending the body. A deny match tells
10675 Squid to delay that control response until the origin server confirms
10676 that the request body is needed. Delaying is the default behavior.
10679 NAME: http_upgrade_request_protocols
10680 TYPE: http_upgrade_request_protocols
10681 LOC: Config.http_upgrade_request_protocols
10683 DEFAULT_DOC: Upgrade header dropped, effectively blocking an upgrade attempt.
10685 Controls client-initiated and server-confirmed switching from HTTP to
10686 another protocol (or to several protocols) using HTTP Upgrade mechanism
10687 defined in RFC 7230 Section 6.7. Squid itself does not understand the
10688 protocols being upgraded to and participates in the upgraded
10689 communication only as a dumb TCP proxy. Admins should not allow
10690 upgrading to protocols that require a more meaningful proxy
10693 Usage: http_upgrade_request_protocols <protocol> allow|deny [!]acl ...
10695 The required "protocol" parameter is either an all-caps word OTHER or an
10696 explicit protocol name (e.g. "WebSocket") optionally followed by a slash
10697 and a version token (e.g. "HTTP/3"). Explicit protocol names and
10698 versions are case sensitive.
10700 When an HTTP client sends an Upgrade request header, Squid iterates over
10701 the client-offered protocols and, for each protocol P (with an optional
10702 version V), evaluates the first non-empty set of
10703 http_upgrade_request_protocols rules (if any) from the following list:
10705 * All rules with an explicit protocol name equal to P.
10706 * All rules that use OTHER instead of a protocol name.
10708 In other words, rules using OTHER are considered for protocol P if and
10709 only if there are no rules mentioning P by name.
10711 If both of the above sets are empty, then Squid removes protocol P from
10714 If the client sent a versioned protocol offer P/X, then explicit rules
10715 referring to the same-name but different-version protocol P/Y are
10716 declared inapplicable. Inapplicable rules are not evaluated (i.e. are
10717 ignored). However, inapplicable rules still belong to the first set of
10720 Within the applicable rule subset, individual rules are evaluated in
10721 their configuration order. If all ACLs of an applicable "allow" rule
10722 match, then the protocol offered by the client is forwarded to the next
10723 hop as is. If all ACLs of an applicable "deny" rule match, then the
10724 offer is dropped. If no applicable rules have matching ACLs, then the
10725 offer is also dropped. The first matching rule also ends rules
10726 evaluation for the offered protocol.
10728 If all client-offered protocols are removed, then Squid forwards the
10729 client request without the Upgrade header. Squid never sends an empty
10730 Upgrade request header.
10732 An Upgrade request header with a value violating HTTP syntax is dropped
10733 and ignored without an attempt to use extractable individual protocol
10736 Upon receiving an HTTP 101 (Switching Protocols) control message, Squid
10737 checks that the server listed at least one protocol name and sent a
10738 Connection:upgrade response header. Squid does not understand individual
10739 protocol naming and versioning concepts enough to implement stricter
10740 checks, but an admin can restrict HTTP 101 (Switching Protocols)
10741 responses further using http_reply_access. Responses denied by
10742 http_reply_access rules and responses flagged by the internal Upgrade
10743 checks result in HTTP 502 (Bad Gateway) ERR_INVALID_RESP errors and
10744 Squid-to-server connection closures.
10746 If Squid sends an Upgrade request header, and the next hop (e.g., the
10747 origin server) responds with an acceptable HTTP 101 (Switching
10748 Protocols), then Squid forwards that message to the client and becomes
10751 The presence of an Upgrade request header alone does not preclude cache
10752 lookups. In other words, an Upgrade request might be satisfied from the
10753 cache, using regular HTTP caching rules.
10755 This clause only supports fast acl types.
10756 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
10758 Each of the following groups of configuration lines represents a
10759 separate configuration example:
10761 # never upgrade to protocol Foo; all others are OK
10762 http_upgrade_request_protocols Foo deny all
10763 http_upgrade_request_protocols OTHER allow all
10765 # only allow upgrades to protocol Bar (except for its first version)
10766 http_upgrade_request_protocols Bar/1 deny all
10767 http_upgrade_request_protocols Bar allow all
10768 http_upgrade_request_protocols OTHER deny all # this rule is optional
10770 # only allow upgrades to protocol Baz, and only if Baz is the only offer
10771 acl UpgradeHeaderHasMultipleOffers ...
10772 http_upgrade_request_protocols Baz deny UpgradeHeaderHasMultipleOffers
10773 http_upgrade_request_protocols Baz allow all
10776 NAME: server_pconn_for_nonretriable
10779 DEFAULT_DOC: Open new connections for forwarding requests Squid cannot retry safely.
10780 LOC: Config.accessList.serverPconnForNonretriable
10782 This option provides fine-grained control over persistent connection
10783 reuse when forwarding HTTP requests that Squid cannot retry. It is useful
10784 in environments where opening new connections is very expensive
10785 (e.g., all connections are secured with TLS with complex client and server
10786 certificate validation) and race conditions associated with persistent
10787 connections are very rare and/or only cause minor problems.
10789 HTTP prohibits retrying unsafe and non-idempotent requests (e.g., POST).
10790 Squid limitations also prohibit retrying all requests with bodies (e.g., PUT).
10791 By default, when forwarding such "risky" requests, Squid opens a new
10792 connection to the server or cache_peer, even if there is an idle persistent
10793 connection available. When Squid is configured to risk sending a non-retriable
10794 request on a previously used persistent connection, and the server closes
10795 the connection before seeing that risky request, the user gets an error response
10796 from Squid. In most cases, that error response will be HTTP 502 (Bad Gateway)
10797 with ERR_ZERO_SIZE_OBJECT or ERR_WRITE_ERROR (peer connection reset) error detail.
10799 If an allow rule matches, Squid reuses an available idle persistent connection
10800 (if any) for the request that Squid cannot retry. If a deny rule matches, then
10801 Squid opens a new connection for the request that Squid cannot retry.
10803 This option does not affect requests that Squid can retry. They will reuse idle
10804 persistent connections (if any).
10806 This clause only supports fast acl types.
10807 See https://wiki.squid-cache.org/SquidFaq/SquidAcl for details.
10810 acl SpeedIsWorthTheRisk method POST
10811 server_pconn_for_nonretriable allow SpeedIsWorthTheRisk
10814 NAME: happy_eyeballs_connect_timeout
10818 LOC: Config.happyEyeballs.connect_timeout
10820 This Happy Eyeballs (RFC 8305) tuning directive specifies the minimum
10821 delay between opening a primary to-server connection and opening a
10822 spare to-server connection for the same master transaction. This delay
10823 is similar to the Connection Attempt Delay in RFC 8305, but it is only
10824 applied to the first spare connection attempt. Subsequent spare
10825 connection attempts use happy_eyeballs_connect_gap, and primary
10826 connection attempts are not artificially delayed at all.
10828 Terminology: The "primary" and "spare" designations are determined by
10829 the order of DNS answers received by Squid: If Squid DNS AAAA query
10830 was answered first, then primary connections are connections to IPv6
10831 peer addresses (while spare connections use IPv4 addresses).
10832 Similarly, if Squid DNS A query was answered first, then primary
10833 connections are connections to IPv4 peer addresses (while spare
10834 connections use IPv6 addresses).
10836 Shorter happy_eyeballs_connect_timeout values reduce master
10837 transaction response time, potentially improving user-perceived
10838 response times (i.e., making user eyeballs happier). Longer delays
10839 reduce both concurrent connection level and server bombardment with
10840 connection requests, potentially improving overall Squid performance
10841 and reducing the chance of being blocked by servers for opening too
10842 many unused connections.
10844 RFC 8305 prohibits happy_eyeballs_connect_timeout values smaller than
10845 10 (milliseconds) to "avoid congestion collapse in the presence of
10846 high packet-loss rates".
10848 The following Happy Eyeballs directives place additional connection
10849 opening restrictions: happy_eyeballs_connect_gap and
10850 happy_eyeballs_connect_limit.
10853 NAME: happy_eyeballs_connect_gap
10857 DEFAULT_DOC: no artificial delays between spare attempts
10858 LOC: Config.happyEyeballs.connect_gap
10860 This Happy Eyeballs (RFC 8305) tuning directive specifies the
10861 minimum delay between opening spare to-server connections (to any
10862 server; i.e. across all concurrent master transactions in a Squid
10863 instance). Each SMP worker currently multiplies the configured gap
10864 by the total number of workers so that the combined spare connection
10865 opening rate of a Squid instance obeys the configured limit. The
10866 workers do not coordinate connection openings yet; a micro burst
10867 of spare connection openings may violate the configured gap.
10869 This directive has similar trade-offs as
10870 happy_eyeballs_connect_timeout, but its focus is on limiting traffic
10871 amplification effects for Squid as a whole, while
10872 happy_eyeballs_connect_timeout works on an individual master
10875 The following Happy Eyeballs directives place additional connection
10876 opening restrictions: happy_eyeballs_connect_timeout and
10877 happy_eyeballs_connect_limit. See the former for related terminology.
10880 NAME: happy_eyeballs_connect_limit
10883 DEFAULT_DOC: no artificial limit on the number of concurrent spare attempts
10884 LOC: Config.happyEyeballs.connect_limit
10886 This Happy Eyeballs (RFC 8305) tuning directive specifies the
10887 maximum number of spare to-server connections (to any server; i.e.
10888 across all concurrent master transactions in a Squid instance).
10889 Each SMP worker gets an equal share of the total limit. However,
10890 the workers do not share the actual connection counts yet, so one
10891 (busier) worker cannot "borrow" spare connection slots from another
10892 (less loaded) worker.
10894 Setting this limit to zero disables concurrent use of primary and
10895 spare TCP connections: Spare connection attempts are made only after
10896 all primary attempts fail. However, Squid would still use the
10897 DNS-related optimizations of the Happy Eyeballs approach.
10899 This directive has similar trade-offs as happy_eyeballs_connect_gap,
10900 but its focus is on limiting Squid overheads, while
10901 happy_eyeballs_connect_gap focuses on the origin server and peer
10904 The following Happy Eyeballs directives place additional connection
10905 opening restrictions: happy_eyeballs_connect_timeout and
10906 happy_eyeballs_connect_gap. See the former for related terminology.
projects / thirdparty / squid.git / blob