Syslog: if we can't find the loglevel specified by the configuration string default to 'info' and warn about the unknown configuration option. no way to make syslog totaly silent anymore.
daryder [Fri, 6 Jun 2014 18:13:58 +0000 (14:13 -0400)]
Update ceph.c
Added special case for "JournalWrBytes". This KPI is recorded as a sum/count pair in ceph schema output. But really, we just want the # of bytes changed - so this has been changed to a derive type.
daryder [Fri, 30 May 2014 19:35:05 +0000 (15:35 -0400)]
Update ceph.c
Changed latency from long run average to current_val-last_val/cur_time-last_time. Also changed counter types to derive types with min = 0, to prevent large data on ceph service restart
Benjamin Gilbert [Wed, 16 Apr 2014 19:00:11 +0000 (15:00 -0400)]
lvm: Report used/free space for thin pools and thick snapshots
Thin pools and traditional snapshot volumes fill up as I/O is performed
to their associated thin volumes or origins. When they run out of free
space, things break, so it's useful to monitor their disk utilization.
Add new plugin instances for thin pool data LVs, thin pool metadata LVs,
and thick snapshot LVs, each with "used" and "free" type instances.
Benjamin Gilbert [Wed, 16 Apr 2014 13:35:56 +0000 (09:35 -0400)]
lvm: Ignore virtual volumes
The sum of the sizes reported by a volume group should equal the size
of the volume group. Virtual volumes do not directly correspond to
allocated storage (and, in fact, may be larger than the entire volume
group), so must be ignored.
Jan Kundrát [Tue, 11 Mar 2014 10:00:22 +0000 (11:00 +0100)]
thresholds: Fix calculation of hysteresis
The old code would never emit a notification when the value changed between the
WARNING and FAILURE; the reason was that each branch in the switch statement
only checked the "old" thresholds valid for the previously encountered state.
That is wrong and pretty dangerous, as there will be no notification when a
values progresses slowly from OK to WARN to FAIL.
write_riemann plugin: Receive acknowledge message when using TCP.
Not receiving an acknowledge message when communicating with riemann over TCP
will cause the riemann instance to eventually hang for extended periods of time
because of resource exhaustion.
Took the time to reaorganize the riemann_send function to simplify locking.
Jan Kundrát [Wed, 26 Feb 2014 14:45:01 +0000 (15:45 +0100)]
Add support for monitoring lm_sensors' "power" sensors
The patch shall probably be extended to include additional sensor names in
future. My servers, however, use "power1" for all of the readings, so I'm not
making that change now.
Thomas D [Wed, 12 Feb 2014 14:32:15 +0000 (15:32 +0100)]
Make sure that "PIDFile" option cannot be set when command-line option "-P" was used. (Fixes #553)
The "PIDFile" value from command-line option "-P", which should take
precedence over any "PIDFile" value from configuration files, could be
overwritten if a configuration file contained the "PIDFile" option.
We introduced a new global variable "pidfile_from_cli" which will be set
to 1 when the command-line option "-P" was detected and the
"global_option_set" function will only set the "PIDFile" option if
"pidfile_from_cli" is 0.
Marc Fournier [Fri, 7 Feb 2014 13:51:31 +0000 (14:51 +0100)]
zfs_arc: support for zfsonlinux
The plugin is structured for Solaris & FreeBSD's get_kstat_value() &
sysctlbyname() functions, which return a single metric based on its
name. ZFS-on-Linux publishes all the stats together in a file in /proc,
which is quite a different approach.
So the idea is to simulate the Solaris/FreeBSD way of doing by using
llist_search() on a linked-list containing all the key/values from the
/proc file. This prevents having to parse it over and over again for
each metric we request.
Patrick Mooney [Mon, 10 Feb 2014 16:00:15 +0000 (10:00 -0600)]
disk: Add udev-based device renaming on Linux
On systems with large collections of allocated disk resources, the
kernel provided names can be difficult to use to logically group or
compare collected values. To grant users the ability to assign custom
instance names for disk devices, collectd can query for a specified udev
attribute. If the attribute is present, its value is subsituted for the
kernel name when submitting the data points.
Marc Fournier [Mon, 3 Feb 2014 00:08:00 +0000 (01:08 +0100)]
auto{conf,make} kernel-related macros consistency
- add AC_DEFINEs for 2 kernels which were missing them.
- declare AM_CONDITIONALs for all kernels we know about + move
declaration at a more sensible location in the file.
Sebastian Harl [Wed, 22 Jan 2014 12:34:30 +0000 (13:34 +0100)]
configure: Fixed linker options when checking for amqp_tcp_socket_new.
This fixes the check when using GCC's --as-needed which fails when options are
not specified in the right order. Libraries (-l options) have to be specified
after any source files. autoconf does the right thing when using $LIBS for
that case rather than $LDFLAGS.