1 .\" Copyright Neil Brown and others.
2 .\" This program is free software; you can redistribute it and/or modify
3 .\" it under the terms of the GNU General Public License as published by
4 .\" the Free Software Foundation; either version 2 of the License, or
5 .\" (at your option) any later version.
6 .\" See file COPYING in distribution for details.
10 md \- Multiple Device driver aka Linux Software RAID
20 driver provides virtual devices that are created from one or more
21 independent underlying devices. This array of devices often contains
22 redundancy and the devices are often disk drives, hence the acronym RAID
23 which stands for a Redundant Array of Independent Disks.
28 4 (striped array with parity device),
29 5 (striped array with distributed parity information),
30 6 (striped array with distributed dual redundancy information), and
31 10 (striped and mirrored).
32 If some number of underlying devices fails while using one of these
33 levels, the array will continue to function; this number is one for
34 RAID levels 4 and 5, two for RAID level 6, and all but one (N-1) for
35 RAID level 1, and dependent on configuration for level 10.
38 also supports a number of pseudo RAID (non-redundant) configurations
39 including RAID0 (striped array), LINEAR (catenated array),
40 MULTIPATH (a set of different interfaces to the same device),
41 and FAULTY (a layer over a single device into which errors can be injected).
44 Each device in an array may have some
46 stored in the device. This metadata is sometimes called a
48 The metadata records information about the structure and state of the array.
49 This allows the array to be reliably re-assembled after a shutdown.
51 From Linux kernel version 2.6.10,
53 provides support for two different formats of metadata, and
54 other formats can be added. Prior to this release, only one format is
57 The common format \(em known as version 0.90 \(em has
58 a superblock that is 4K long and is written into a 64K aligned block that
59 starts at least 64K and less than 128K from the end of the device
60 (i.e. to get the address of the superblock round the size of the
61 device down to a multiple of 64K and then subtract 64K).
62 The available size of each device is the amount of space before the
63 super block, so between 64K and 128K is lost when a device in
64 incorporated into an MD array.
65 This superblock stores multi-byte fields in a processor-dependent
66 manner, so arrays cannot easily be moved between computers with
69 The new format \(em known as version 1 \(em has a superblock that is
70 normally 1K long, but can be longer. It is normally stored between 8K
71 and 12K from the end of the device, on a 4K boundary, though
72 variations can be stored at the start of the device (version 1.1) or 4K from
73 the start of the device (version 1.2).
74 This metadata format stores multibyte data in a
75 processor-independent format and supports up to hundreds of
76 component devices (version 0.90 only supports 28).
78 The metadata contains, among other things:
81 The manner in which the devices are arranged into the array
82 (LINEAR, RAID0, RAID1, RAID4, RAID5, RAID10, MULTIPATH).
85 a 128 bit Universally Unique Identifier that identifies the array that
89 When a version 0.90 array is being reshaped (e.g. adding extra devices
90 to a RAID5), the version number is temporarily set to 0.91. This
91 ensures that if the reshape process is stopped in the middle (e.g. by
92 a system crash) and the machine boots into an older kernel that does
93 not support reshaping, then the array will not be assembled (which
94 would cause data corruption) but will be left untouched until a kernel
95 that can complete the reshape processes is used.
97 .SS ARRAYS WITHOUT METADATA
98 While it is usually best to create arrays with superblocks so that
99 they can be assembled reliably, there are some circumstances when an
100 array without superblocks is preferred. These include:
103 Early versions of the
105 driver only supported LINEAR and RAID0 configurations and did not use
106 a superblock (which is less critical with these configurations).
107 While such arrays should be rebuilt with superblocks if possible,
109 continues to support them.
112 Being a largely transparent layer over a different device, the FAULTY
113 personality doesn't gain anything from having a superblock.
116 It is often possible to detect devices which are different paths to
117 the same storage directly rather than having a distinctive superblock
118 written to the device and searched for on all paths. In this case,
119 a MULTIPATH array with no superblock makes sense.
122 In some configurations it might be desired to create a RAID1
123 configuration that does not use a superblock, and to maintain the state of
124 the array elsewhere. While not encouraged for general use, it does
125 have special-purpose uses and is supported.
127 .SS ARRAYS WITH EXTERNAL METADATA
129 From release 2.6.28, the
131 driver supports arrays with externally managed metadata. That is,
132 the metadata is not managed by the kernel but rather by a user-space
133 program which is external to the kernel. This allows support for a
134 variety of metadata formats without cluttering the kernel with lots of
138 is able to communicate with the user-space program through various
139 sysfs attributes so that it can make appropriate changes to the
140 metadata \- for example to mark a device as faulty. When necessary,
142 will wait for the program to acknowledge the event by writing to a
146 contains more detail about this interaction.
149 Many metadata formats use a single block of metadata to describe a
150 number of different arrays which all use the same set of devices.
151 In this case it is helpful for the kernel to know about the full set
152 of devices as a whole. This set is known to md as a
156 array with externally managed metadata and with device offset and size
157 so that it just covers the metadata part of the devices. The
158 remainder of each device is available to be incorporated into various
163 A LINEAR array simply catenates the available space on each
164 drive to form one large virtual drive.
166 One advantage of this arrangement over the more common RAID0
167 arrangement is that the array may be reconfigured at a later time with
168 an extra drive, so the array is made bigger without disturbing the
169 data that is on the array. This can even be done on a live
172 If a chunksize is given with a LINEAR array, the usable space on each
173 device is rounded down to a multiple of this chunksize.
177 A RAID0 array (which has zero redundancy) is also known as a
179 A RAID0 array is configured at creation with a
181 which must be a power of two (prior to Linux 2.6.31), and at least 4
184 The RAID0 driver assigns the first chunk of the array to the first
185 device, the second chunk to the second device, and so on until all
186 drives have been assigned one chunk. This collection of chunks forms a
188 Further chunks are gathered into stripes in the same way, and are
189 assigned to the remaining space in the drives.
191 If devices in the array are not all the same size, then once the
192 smallest device has been exhausted, the RAID0 driver starts
193 collecting chunks into smaller stripes that only span the drives which
194 still have remaining space.
196 A bug was introduced in linux 3.14 which changed the layout of blocks in
197 a RAID0 beyond the region that is striped over all devices. This bug
198 does not affect an array with all devices the same size, but can affect
201 Linux 5.4 (and some stable kernels to which the change was backported)
202 will not normally assemble such an array as it cannot know which layout
203 to use. There is a module parameter "raid0.default_layout" which can be
204 set to "1" to force the kernel to use the pre-3.14 layout or to "2" to
205 force it to use the 3.14-and-later layout. when creating a new RAID0
208 will record the chosen layout in the metadata in a way that allows newer
209 kernels to assemble the array without needing a module parameter.
211 To assemble an old array on a new kernel without using the module parameter,
213 .B "--update=layout-original"
215 .B "--update=layout-alternate"
218 Once you have updated the layout you will not be able to mount the array
219 on an older kernel. If you need to revert to an older kernel, the
220 layout information can be erased with the
221 .B "--update=layout-unspecificed"
222 option. If you use this option to
224 while running a newer kernel, the array will NOT assemble, but the
225 metadata will be update so that it can be assembled on an older kernel.
227 No that setting the layout to "unspecified" removes protections against
228 this bug, and you must be sure that the kernel you use matches the
233 A RAID1 array is also known as a mirrored set (though mirrors tend to
234 provide reflected images, which RAID1 does not) or a plex.
236 Once initialised, each device in a RAID1 array contains exactly the
237 same data. Changes are written to all devices in parallel. Data is
238 read from any one device. The driver attempts to distribute read
239 requests across all devices to maximise performance.
241 All devices in a RAID1 array should be the same size. If they are
242 not, then only the amount of space available on the smallest device is
243 used (any extra space on other devices is wasted).
245 Note that the read balancing done by the driver does not make the RAID1
246 performance profile be the same as for RAID0; a single stream of
247 sequential input will not be accelerated (e.g. a single dd), but
248 multiple sequential streams or a random workload will use more than one
249 spindle. In theory, having an N-disk RAID1 will allow N sequential
250 threads to read from all disks.
252 Individual devices in a RAID1 can be marked as "write-mostly".
253 These drives are excluded from the normal read balancing and will only
254 be read from when there is no other option. This can be useful for
255 devices connected over a slow link.
259 A RAID4 array is like a RAID0 array with an extra device for storing
260 parity. This device is the last of the active devices in the
261 array. Unlike RAID0, RAID4 also requires that all stripes span all
262 drives, so extra space on devices that are larger than the smallest is
265 When any block in a RAID4 array is modified, the parity block for that
266 stripe (i.e. the block in the parity device at the same device offset
267 as the stripe) is also modified so that the parity block always
268 contains the "parity" for the whole stripe. I.e. its content is
269 equivalent to the result of performing an exclusive-or operation
270 between all the data blocks in the stripe.
272 This allows the array to continue to function if one device fails.
273 The data that was on that device can be calculated as needed from the
274 parity block and the other data blocks.
278 RAID5 is very similar to RAID4. The difference is that the parity
279 blocks for each stripe, instead of being on a single device, are
280 distributed across all devices. This allows more parallelism when
281 writing, as two different block updates will quite possibly affect
282 parity blocks on different devices so there is less contention.
284 This also allows more parallelism when reading, as read requests are
285 distributed over all the devices in the array instead of all but one.
289 RAID6 is similar to RAID5, but can handle the loss of any \fItwo\fP
290 devices without data loss. Accordingly, it requires N+2 drives to
291 store N drives worth of data.
293 The performance for RAID6 is slightly lower but comparable to RAID5 in
294 normal mode and single disk failure mode. It is very slow in dual
295 disk failure mode, however.
299 RAID10 provides a combination of RAID1 and RAID0, and is sometimes known
300 as RAID1+0. Every datablock is duplicated some number of times, and
301 the resulting collection of datablocks are distributed over multiple
304 When configuring a RAID10 array, it is necessary to specify the number
305 of replicas of each data block that are required (this will usually
306 be\ 2) and whether their layout should be "near", "far" or "offset"
307 (with "offset" being available since Linux\ 2.6.18).
309 .B About the RAID10 Layout Examples:
311 The examples below visualise the chunk distribution on the underlying
312 devices for the respective layout.
314 For simplicity it is assumed that the size of the chunks equals the
315 size of the blocks of the underlying devices as well as those of the
316 RAID10 device exported by the kernel (for example \fB/dev/md/\fPname).
318 Therefore the chunks\ /\ chunk numbers map directly to the blocks\ /\
319 block addresses of the exported RAID10 device.
321 Decimal numbers (0,\ 1, 2,\ ...) are the chunks of the RAID10 and due
322 to the above assumption also the blocks and block addresses of the
323 exported RAID10 device.
325 Repeated numbers mean copies of a chunk\ /\ block (obviously on
326 different underlying devices).
328 Hexadecimal numbers (0x00,\ 0x01, 0x02,\ ...) are the block addresses
329 of the underlying devices.
333 When "near" replicas are chosen, the multiple copies of a given chunk are laid
334 out consecutively ("as close to each other as possible") across the stripes of
337 With an even number of devices, they will likely (unless some misalignment is
338 present) lay at the very same offset on the different devices.
340 This is as the "classic" RAID1+0; that is two groups of mirrored devices (in the
341 example below the groups Device\ #1\ /\ #2 and Device\ #3\ /\ #4 are each a
342 RAID1) both in turn forming a striped RAID0.
345 .B Example with 2\ copies per chunk and an even number\ (4) of devices:
350 | - | - | - | - | - |
351 | C | C | C | C | C |
352 | C | C | C | C | C |
353 | C | C | C | C | C |
354 | C | C | C | C | C |
355 | C | C | C | C | C |
356 | C | C | C | C | C |
357 | - | - | - | - | - |
363 ;Device #1;Device #2;Device #3;Device #4
366 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.
368 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.
370 ;\\---------v---------/;\\---------v---------/
372 ;\\---------------------v---------------------/
377 .B Example with 2\ copies per chunk and an odd number\ (5) of devices:
381 C | C | C | C | C | C |
382 | - | - | - | - | - | - |
383 | C | C | C | C | C | C |
384 | C | C | C | C | C | C |
385 | C | C | C | C | C | C |
386 | C | C | C | C | C | C |
387 | C | C | C | C | C | C |
388 | C | C | C | C | C | C |
389 | - | - | - | - | - | - |
392 ;Dev #1;Dev #2;Dev #3;Dev #4;Dev #5
395 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.
397 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.
398 0x80;317;318;318;319;319
404 When "far" replicas are chosen, the multiple copies of a given chunk
405 are laid out quite distant ("as far as reasonably possible") from each
408 First a complete sequence of all data blocks (that is all the data one
409 sees on the exported RAID10 block device) is striped over the
410 devices. Then another (though "shifted") complete sequence of all data
411 blocks; and so on (in the case of more than 2\ copies per chunk).
413 The "shift" needed to prevent placing copies of the same chunks on the
414 same devices is actually a cyclic permutation with offset\ 1 of each
415 of the stripes within a complete sequence of chunks.
417 The offset\ 1 is relative to the previous complete sequence of chunks,
418 so in case of more than 2\ copies per chunk one gets the following
421 1.\ complete sequence of chunks: offset\ =\ \ 0
423 2.\ complete sequence of chunks: offset\ =\ \ 1
425 3.\ complete sequence of chunks: offset\ =\ \ 2
429 n.\ complete sequence of chunks: offset\ =\ n-1
432 .B Example with 2\ copies per chunk and an even number\ (4) of devices:
437 | - | - | - | - | - |
438 | C | C | C | C | C | L
439 | C | C | C | C | C | L
440 | C | C | C | C | C | L
441 | C | C | C | C | C | L
442 | C | C | C | C | C | L
443 | C | C | C | C | C | L
444 | C | C | C | C | C | L
445 | C | C | C | C | C | L
446 | C | C | C | C | C | L
447 | C | C | C | C | C | L
448 | C | C | C | C | C | L
449 | C | C | C | C | C | L
450 | - | - | - | - | - |
453 ;Device #1;Device #2;Device #3;Device #4
457 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
459 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
460 0x40;252;253;254;255;/
463 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
465 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
466 0x80;255;252;253;254;/
471 .B Example with 2\ copies per chunk and an odd number\ (5) of devices:
475 C | C | C | C | C | C |
476 | - | - | - | - | - | - |
477 | C | C | C | C | C | C | L
478 | C | C | C | C | C | C | L
479 | C | C | C | C | C | C | L
480 | C | C | C | C | C | C | L
481 | C | C | C | C | C | C | L
482 | C | C | C | C | C | C | L
483 | C | C | C | C | C | C | L
484 | C | C | C | C | C | C | L
485 | C | C | C | C | C | C | L
486 | C | C | C | C | C | C | L
487 | C | C | C | C | C | C | L
488 | C | C | C | C | C | C | L
489 | - | - | - | - | - | - |
492 ;Dev #1;Dev #2;Dev #3;Dev #4;Dev #5
496 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
498 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
499 0x40;315;316;317;318;319;/
501 0x42;9;5;6;7;8;> [#]~
502 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
504 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
505 0x80;319;315;316;317;318;/
509 With [#]\ being the complete sequence of chunks and [#]~\ the cyclic permutation
510 with offset\ 1 thereof (in the case of more than 2 copies per chunk there would
511 be ([#]~)~,\ (([#]~)~)~,\ ...).
513 The advantage of this layout is that MD can easily spread sequential reads over
514 the devices, making them similar to RAID0 in terms of speed.
516 The cost is more seeking for writes, making them substantially slower.
519 \fB"offset" Layout\fP
520 When "offset" replicas are chosen, all the copies of a given chunk are
521 striped consecutively ("offset by the stripe length after each other")
524 Explained in detail, <number of devices> consecutive chunks are
525 striped over the devices, immediately followed by a "shifted" copy of
526 these chunks (and by further such "shifted" copies in the case of more
527 than 2\ copies per chunk).
529 This pattern repeats for all further consecutive chunks of the
530 exported RAID10 device (in other words: all further data blocks).
532 The "shift" needed to prevent placing copies of the same chunks on the
533 same devices is actually a cyclic permutation with offset\ 1 of each
534 of the striped copies of <number of devices> consecutive chunks.
536 The offset\ 1 is relative to the previous striped copy of <number of
537 devices> consecutive chunks, so in case of more than 2\ copies per
538 chunk one gets the following offsets:
540 1.\ <number of devices> consecutive chunks: offset\ =\ \ 0
542 2.\ <number of devices> consecutive chunks: offset\ =\ \ 1
544 3.\ <number of devices> consecutive chunks: offset\ =\ \ 2
548 n.\ <number of devices> consecutive chunks: offset\ =\ n-1
551 .B Example with 2\ copies per chunk and an even number\ (4) of devices:
556 | - | - | - | - | - |
557 | C | C | C | C | C | L
558 | C | C | C | C | C | L
559 | C | C | C | C | C | L
560 | C | C | C | C | C | L
561 | C | C | C | C | C | L
562 | C | C | C | C | C | L
563 | C | C | C | C | C | L
564 | C | C | C | C | C | L
565 | C | C | C | C | C | L
566 | - | - | - | - | - |
569 ;Device #1;Device #2;Device #3;Device #4
575 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;) \.\.\.
577 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;) \.\.\.
578 0x79;251;252;253;254;) EX
579 0x80;254;251;252;253;) EX~
584 .B Example with 2\ copies per chunk and an odd number\ (5) of devices:
588 C | C | C | C | C | C |
589 | - | - | - | - | - | - |
590 | C | C | C | C | C | C | L
591 | C | C | C | C | C | C | L
592 | C | C | C | C | C | C | L
593 | C | C | C | C | C | C | L
594 | C | C | C | C | C | C | L
595 | C | C | C | C | C | C | L
596 | C | C | C | C | C | C | L
597 | C | C | C | C | C | C | L
598 | C | C | C | C | C | C | L
599 | - | - | - | - | - | - |
602 ;Dev #1;Dev #2;Dev #3;Dev #4;Dev #5
608 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;) \.\.\.
610 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;) \.\.\.
611 0x79;314;315;316;317;318;) EX
612 0x80;318;314;315;316;317;) EX~
616 With AA,\ AB,\ ..., AZ,\ BA,\ ... being the sets of <number of devices> consecutive
617 chunks and AA~,\ AB~,\ ..., AZ~,\ BA~,\ ... the cyclic permutations with offset\ 1
618 thereof (in the case of more than 2 copies per chunk there would be (AA~)~,\ ...
619 as well as ((AA~)~)~,\ ... and so on).
621 This should give similar read characteristics to "far" if a suitably large chunk
622 size is used, but without as much seeking for writes.
626 It should be noted that the number of devices in a RAID10 array need
627 not be a multiple of the number of replica of each data block; however,
628 there must be at least as many devices as replicas.
630 If, for example, an array is created with 5 devices and 2 replicas,
631 then space equivalent to 2.5 of the devices will be available, and
632 every block will be stored on two different devices.
634 Finally, it is possible to have an array with both "near" and "far"
635 copies. If an array is configured with 2 near copies and 2 far
636 copies, then there will be a total of 4 copies of each block, each on
637 a different drive. This is an artifact of the implementation and is
638 unlikely to be of real value.
642 MULTIPATH is not really a RAID at all as there is only one real device
643 in a MULTIPATH md array. However there are multiple access points
644 (paths) to this device, and one of these paths might fail, so there
645 are some similarities.
647 A MULTIPATH array is composed of a number of logically different
648 devices, often fibre channel interfaces, that all refer the the same
649 real device. If one of these interfaces fails (e.g. due to cable
650 problems), the MULTIPATH driver will attempt to redirect requests to
653 The MULTIPATH drive is not receiving any ongoing development and
654 should be considered a legacy driver. The device-mapper based
655 multipath drivers should be preferred for new installations.
658 The FAULTY md module is provided for testing purposes. A FAULTY array
659 has exactly one component device and is normally assembled without a
660 superblock, so the md array created provides direct access to all of
661 the data in the component device.
663 The FAULTY module may be requested to simulate faults to allow testing
664 of other md levels or of filesystems. Faults can be chosen to trigger
665 on read requests or write requests, and can be transient (a subsequent
666 read/write at the address will probably succeed) or persistent
667 (subsequent read/write of the same address will fail). Further, read
668 faults can be "fixable" meaning that they persist until a write
669 request at the same address.
671 Fault types can be requested with a period. In this case, the fault
672 will recur repeatedly after the given number of requests of the
673 relevant type. For example if persistent read faults have a period of
674 100, then every 100th read request would generate a fault, and the
675 faulty sector would be recorded so that subsequent reads on that
676 sector would also fail.
678 There is a limit to the number of faulty sectors that are remembered.
679 Faults generated after this limit is exhausted are treated as
682 The list of faulty sectors can be flushed, and the active list of
683 failure modes can be cleared.
687 When changes are made to a RAID1, RAID4, RAID5, RAID6, or RAID10 array
688 there is a possibility of inconsistency for short periods of time as
689 each update requires at least two block to be written to different
690 devices, and these writes probably won't happen at exactly the same
691 time. Thus if a system with one of these arrays is shutdown in the
692 middle of a write operation (e.g. due to power failure), the array may
695 To handle this situation, the md driver marks an array as "dirty"
696 before writing any data to it, and marks it as "clean" when the array
697 is being disabled, e.g. at shutdown. If the md driver finds an array
698 to be dirty at startup, it proceeds to correct any possibly
699 inconsistency. For RAID1, this involves copying the contents of the
700 first drive onto all other drives. For RAID4, RAID5 and RAID6 this
701 involves recalculating the parity for each stripe and making sure that
702 the parity block has the correct data. For RAID10 it involves copying
703 one of the replicas of each block onto all the others. This process,
704 known as "resynchronising" or "resync" is performed in the background.
705 The array can still be used, though possibly with reduced performance.
707 If a RAID4, RAID5 or RAID6 array is degraded (missing at least one
708 drive, two for RAID6) when it is restarted after an unclean shutdown, it cannot
709 recalculate parity, and so it is possible that data might be
710 undetectably corrupted. The 2.4 md driver
712 alert the operator to this condition. The 2.6 md driver will fail to
713 start an array in this condition without manual intervention, though
714 this behaviour can be overridden by a kernel parameter.
718 If the md driver detects a write error on a device in a RAID1, RAID4,
719 RAID5, RAID6, or RAID10 array, it immediately disables that device
720 (marking it as faulty) and continues operation on the remaining
721 devices. If there are spare drives, the driver will start recreating
722 on one of the spare drives the data which was on that failed drive,
723 either by copying a working drive in a RAID1 configuration, or by
724 doing calculations with the parity block on RAID4, RAID5 or RAID6, or
725 by finding and copying originals for RAID10.
727 In kernels prior to about 2.6.15, a read error would cause the same
728 effect as a write error. In later kernels, a read-error will instead
729 cause md to attempt a recovery by overwriting the bad block. i.e. it
730 will find the correct data from elsewhere, write it over the block
731 that failed, and then try to read it back again. If either the write
732 or the re-read fail, md will treat the error the same way that a write
733 error is treated, and will fail the whole device.
735 While this recovery process is happening, the md driver will monitor
736 accesses to the array and will slow down the rate of recovery if other
737 activity is happening, so that normal access to the array will not be
738 unduly affected. When no other activity is happening, the recovery
739 process proceeds at full speed. The actual speed targets for the two
740 different situations can be controlled by the
744 control files mentioned below.
746 .SS SCRUBBING AND MISMATCHES
748 As storage devices can develop bad blocks at any time it is valuable
749 to regularly read all blocks on all devices in an array so as to catch
750 such bad blocks early. This process is called
753 md arrays can be scrubbed by writing either
761 directory for the device.
763 Requesting a scrub will cause
765 to read every block on every device in the array, and check that the
766 data is consistent. For RAID1 and RAID10, this means checking that the copies
767 are identical. For RAID4, RAID5, RAID6 this means checking that the
768 parity block is (or blocks are) correct.
770 If a read error is detected during this process, the normal read-error
771 handling causes correct data to be found from other devices and to be
772 written back to the faulty device. In many case this will
777 If all blocks read successfully but are found to not be consistent,
778 then this is regarded as a
783 was used, then no action is taken to handle the mismatch, it is simply
787 was used, then a mismatch will be repaired in the same way that
789 repairs arrays. For RAID5/RAID6 new parity blocks are written. For RAID1/RAID10,
790 all but one block are overwritten with the content of that one block.
792 A count of mismatches is recorded in the
795 .IR md/mismatch_cnt .
796 This is set to zero when a
797 scrub starts and is incremented whenever a sector is
798 found that is a mismatch.
800 normally works in units much larger than a single sector and when it
801 finds a mismatch, it does not determine exactly how many actual sectors were
802 affected but simply adds the number of sectors in the IO unit that was
803 used. So a value of 128 could simply mean that a single 64KB check
804 found an error (128 x 512bytes = 64KB).
806 If an array is created by
810 then a subsequent check could be expected to find some mismatches.
812 On a truly clean RAID5 or RAID6 array, any mismatches should indicate
813 a hardware problem at some level - software issues should never cause
816 However on RAID1 and RAID10 it is possible for software issues to
817 cause a mismatch to be reported. This does not necessarily mean that
818 the data on the array is corrupted. It could simply be that the
819 system does not care what is stored on that part of the array - it is
822 The most likely cause for an unexpected mismatch on RAID1 or RAID10
823 occurs if a swap partition or swap file is stored on the array.
825 When the swap subsystem wants to write a page of memory out, it flags
826 the page as 'clean' in the memory manager and requests the swap device
827 to write it out. It is quite possible that the memory will be
828 changed while the write-out is happening. In that case the 'clean'
829 flag will be found to be clear when the write completes and so the
830 swap subsystem will simply forget that the swapout had been attempted,
831 and will possibly choose a different page to write out.
833 If the swap device was on RAID1 (or RAID10), then the data is sent
834 from memory to a device twice (or more depending on the number of
835 devices in the array). Thus it is possible that the memory gets changed
836 between the times it is sent, so different data can be written to
837 the different devices in the array. This will be detected by
839 as a mismatch. However it does not reflect any corruption as the
840 block where this mismatch occurs is being treated by the swap system as
841 being empty, and the data will never be read from that block.
843 It is conceivable for a similar situation to occur on non-swap files,
844 though it is less likely.
848 value can not be interpreted very reliably on RAID1 or RAID10,
849 especially when the device is used for swap.
852 .SS BITMAP WRITE-INTENT LOGGING
856 supports a bitmap based write-intent log. If configured, the bitmap
857 is used to record which blocks of the array may be out of sync.
858 Before any write request is honoured, md will make sure that the
859 corresponding bit in the log is set. After a period of time with no
860 writes to an area of the array, the corresponding bit will be cleared.
862 This bitmap is used for two optimisations.
864 Firstly, after an unclean shutdown, the resync process will consult
865 the bitmap and only resync those blocks that correspond to bits in the
866 bitmap that are set. This can dramatically reduce resync time.
868 Secondly, when a drive fails and is removed from the array, md stops
869 clearing bits in the intent log. If that same drive is re-added to
870 the array, md will notice and will only recover the sections of the
871 drive that are covered by bits in the intent log that are set. This
872 can allow a device to be temporarily removed and reinserted without
873 causing an enormous recovery cost.
875 The intent log can be stored in a file on a separate device, or it can
876 be stored near the superblocks of an array which has superblocks.
878 It is possible to add an intent log to an active array, or remove an
879 intent log if one is present.
881 In 2.6.13, intent bitmaps are only supported with RAID1. Other levels
882 with redundancy are supported from 2.6.15.
886 From Linux 3.5 each device in an
888 array can store a list of known-bad-blocks. This list is 4K in size
889 and usually positioned at the end of the space between the superblock
892 When a block cannot be read and cannot be repaired by writing data
893 recovered from other devices, the address of the block is stored in
894 the bad block list. Similarly if an attempt to write a block fails,
895 the address will be recorded as a bad block. If attempting to record
896 the bad block fails, the whole device will be marked faulty.
898 Attempting to read from a known bad block will cause a read error.
899 Attempting to write to a known bad block will be ignored if any write
900 errors have been reported by the device. If there have been no write
901 errors then the data will be written to the known bad block and if
902 that succeeds, the address will be removed from the list.
904 This allows an array to fail more gracefully - a few blocks on
905 different devices can be faulty without taking the whole array out of
908 The list is particularly useful when recovering to a spare. If a few blocks
909 cannot be read from the other devices, the bulk of the recovery can
910 complete and those few bad blocks will be recorded in the bad block list.
912 .SS RAID456 WRITE JOURNAL
914 Due to non-atomicity nature of RAID write operations, interruption of
915 write operations (system crash, etc.) to RAID456 array can lead to
916 inconsistent parity and data loss (so called RAID-5 write hole).
918 To plug the write hole, from Linux 4.4 (to be confirmed),
920 supports write ahead journal for RAID456. When the array is created,
921 an additional journal device can be added to the array through
923 option. The RAID write journal works similar to file system journals.
924 Before writing to the data disks, md persists data AND parity of the
925 stripe to the journal device. After crashes, md searches the journal
926 device for incomplete write operations, and replay them to the data
929 When the journal device fails, the RAID array is forced to run in
936 supports WRITE-BEHIND on RAID1 arrays.
938 This allows certain devices in the array to be flagged as
940 MD will only read from such devices if there is no
943 If a write-intent bitmap is also provided, write requests to
944 write-mostly devices will be treated as write-behind requests and md
945 will not wait for writes to those requests to complete before
946 reporting the write as complete to the filesystem.
948 This allows for a RAID1 with WRITE-BEHIND to be used to mirror data
949 over a slow link to a remote computer (providing the link isn't too
950 slow). The extra latency of the remote link will not slow down normal
951 operations, but the remote system will still have a reasonably
952 up-to-date copy of all data.
959 supports FAILFAST for RAID1 and RAID10 arrays. This is a flag that
960 can be set on individual drives, though it is usually set on all
961 drives, or no drives.
965 sends an I/O request to a drive that is marked as FAILFAST, and when
966 the array could survive the loss of that drive without losing data,
968 will request that the underlying device does not perform any retries.
969 This means that a failure will be reported to
971 promptly, and it can mark the device as faulty and continue using the
974 cannot control the timeout that the underlying devices use to
975 determine failure. Any changes desired to that timeout must be set
976 explictly on the underlying device, separately from using
979 If a FAILFAST request does fail, and if it is still safe to mark the
980 device as faulty without data loss, that will be done and the array
981 will continue functioning on a reduced number of devices. If it is not
982 possible to safely mark the device as faulty,
984 will retry the request without disabling retries in the underlying
987 will not attempt to repair read errors on a device marked as FAILFAST
988 by writing out the correct. It will just mark the device as faulty.
990 FAILFAST is appropriate for storage arrays that have a low probability
991 of true failure, but will sometimes introduce unacceptable delays to
992 I/O requests while performing internal maintenance. The value of
993 setting FAILFAST involves a trade-off. The gain is that the chance of
994 unacceptable delays is substantially reduced. The cost is that the
995 unlikely event of data-loss on one device is slightly more likely to
996 result in data-loss for the array.
998 When a device in an array using FAILFAST is marked as faulty, it will
999 usually become usable again in a short while.
1001 makes no attempt to detect that possibility. Some separate
1002 mechanism, tuned to the specific details of the expected failure modes,
1003 needs to be created to monitor devices to see when they return to full
1004 functionality, and to then re-add them to the array. In order of
1005 this "re-add" functionality to be effective, an array using FAILFAST
1006 should always have a write-intent bitmap.
1013 is the processes of re-arranging the data stored in each stripe into a
1014 new layout. This might involve changing the number of devices in the
1015 array (so the stripes are wider), changing the chunk size (so stripes
1016 are deeper or shallower), or changing the arrangement of data and
1017 parity (possibly changing the RAID level, e.g. 1 to 5 or 5 to 6).
1019 As of Linux 2.6.35, md can reshape a RAID4, RAID5, or RAID6 array to
1020 have a different number of devices (more or fewer) and to have a
1021 different layout or chunk size. It can also convert between these
1022 different RAID levels. It can also convert between RAID0 and RAID10,
1023 and between RAID0 and RAID4 or RAID5.
1024 Other possibilities may follow in future kernels.
1026 During any stripe process there is a 'critical section' during which
1027 live data is being overwritten on disk. For the operation of
1028 increasing the number of drives in a RAID5, this critical section
1029 covers the first few stripes (the number being the product of the old
1030 and new number of devices). After this critical section is passed,
1031 data is only written to areas of the array which no longer hold live
1032 data \(em the live data has already been located away.
1034 For a reshape which reduces the number of devices, the 'critical
1035 section' is at the end of the reshape process.
1037 md is not able to ensure data preservation if there is a crash
1038 (e.g. power failure) during the critical section. If md is asked to
1039 start an array which failed during a critical section of restriping,
1040 it will fail to start the array.
1042 To deal with this possibility, a user-space program must
1044 Disable writes to that section of the array (using the
1048 take a copy of the data somewhere (i.e. make a backup),
1050 allow the process to continue and invalidate the backup and restore
1051 write access once the critical section is passed, and
1053 provide for restoring the critical data before restarting the array
1054 after a system crash.
1058 versions from 2.4 do this for growing a RAID5 array.
1060 For operations that do not change the size of the array, like simply
1061 increasing chunk size, or converting RAID5 to RAID6 with one extra
1062 device, the entire process is the critical section. In this case, the
1063 restripe will need to progress in stages, as a section is suspended,
1064 backed up, restriped, and released.
1067 Each block device appears as a directory in
1069 (which is usually mounted at
1071 For MD devices, this directory will contain a subdirectory called
1073 which contains various files for providing access to information about
1076 This interface is documented more fully in the file
1077 .B Documentation/admin-guide/md.rst
1078 which is distributed with the kernel sources. That file should be
1079 consulted for full documentation. The following are just a selection
1080 of attribute files that are available.
1083 .B md/sync_speed_min
1084 This value, if set, overrides the system-wide setting in
1085 .B /proc/sys/dev/raid/speed_limit_min
1086 for this array only.
1089 to this file will cause the system-wide setting to have effect.
1092 .B md/sync_speed_max
1093 This is the partner of
1094 .B md/sync_speed_min
1096 .B /proc/sys/dev/raid/speed_limit_max
1101 This can be used to monitor and control the resync/recovery process of
1103 In particular, writing "check" here will cause the array to read all
1104 data block and check that they are consistent (e.g. parity is correct,
1105 or all mirror replicas are the same). Any discrepancies found are
1109 A count of problems found will be stored in
1110 .BR md/mismatch_count .
1112 Alternately, "repair" can be written which will cause the same check
1113 to be performed, but any errors will be corrected.
1115 Finally, "idle" can be written to stop the check/repair process.
1118 .B md/stripe_cache_size
1119 This is only available on RAID5 and RAID6. It records the size (in
1120 pages per device) of the stripe cache which is used for synchronising
1121 all write operations to the array and all read operations if the array
1122 is degraded. The default is 256. Valid values are 17 to 32768.
1123 Increasing this number can increase performance in some situations, at
1124 some cost in system memory. Note, setting this value too high can
1125 result in an "out of memory" condition for the system.
1127 memory_consumed = system_page_size * nr_disks * stripe_cache_size
1130 .B md/preread_bypass_threshold
1131 This is only available on RAID5 and RAID6. This variable sets the
1132 number of times MD will service a full-stripe-write before servicing a
1133 stripe that requires some "prereading". For fairness this defaults to
1134 1. Valid values are 0 to stripe_cache_size. Setting this to 0
1135 maximizes sequential-write throughput at the cost of fairness to threads
1136 doing small or random writes.
1139 .B md/bitmap/backlog
1140 The value stored in the file only has any effect on RAID1 when write-mostly
1141 devices are active, and write requests to those devices are proceed in the
1144 This variable sets a limit on the number of concurrent background writes,
1145 the valid values are 0 to 16383, 0 means that write-behind is not allowed,
1146 while any other number means it can happen. If there are more write requests
1147 than the number, new writes will by synchronous.
1150 .B md/bitmap/can_clear
1151 This is for externally managed bitmaps, where the kernel writes the bitmap
1152 itself, but metadata describing the bitmap is managed by mdmon or similar.
1154 When the array is degraded, bits mustn't be cleared. When the array becomes
1155 optimal again, bit can be cleared, but first the metadata needs to record
1156 the current event count. So md sets this to 'false' and notifies mdmon,
1157 then mdmon updates the metadata and writes 'true'.
1159 There is no code in mdmon to actually do this, so maybe it doesn't even
1163 .B md/bitmap/chunksize
1164 The bitmap chunksize can only be changed when no bitmap is active, and
1165 the value should be power of 2 and at least 512.
1168 .B md/bitmap/location
1169 This indicates where the write-intent bitmap for the array is stored.
1170 It can be "none" or "file" or a signed offset from the array metadata
1171 - measured in sectors. You cannot set a file by writing here - that can
1172 only be done with the SET_BITMAP_FILE ioctl.
1174 Write 'none' to 'bitmap/location' will clear bitmap, and the previous
1175 location value must be write to it to restore bitmap.
1178 .B md/bitmap/max_backlog_used
1179 This keeps track of the maximum number of concurrent write-behind requests
1180 for an md array, writing any value to this file will clear it.
1183 .B md/bitmap/metadata
1184 This can be 'internal' or 'clustered' or 'external'. 'internal' is set
1185 by default, which means the metadata for bitmap is stored in the first 256
1186 bytes of the bitmap space. 'clustered' means separate bitmap metadata are
1187 used for each cluster node. 'external' means that bitmap metadata is managed
1188 externally to the kernel.
1192 This shows the space (in sectors) which is available at md/bitmap/location,
1193 and allows the kernel to know when it is safe to resize the bitmap to match
1194 a resized array. It should big enough to contain the total bytes in the bitmap.
1196 For 1.0 metadata, assume we can use up to the superblock if before, else
1197 to 4K beyond superblock. For other metadata versions, assume no change is
1201 .B md/bitmap/time_base
1202 This shows the time (in seconds) between disk flushes, and is used to looking
1203 for bits in the bitmap to be cleared.
1205 The default value is 5 seconds, and it should be an unsigned long value.
1207 .SS KERNEL PARAMETERS
1209 The md driver recognised several different kernel parameters.
1211 .B raid=noautodetect
1212 This will disable the normal detection of md arrays that happens at
1213 boot time. If a drive is partitioned with MS-DOS style partitions,
1214 then if any of the 4 main partitions has a partition type of 0xFD,
1215 then that partition will normally be inspected to see if it is part of
1216 an MD array, and if any full arrays are found, they are started. This
1217 kernel parameter disables this behaviour.
1220 .B raid=partitionable
1223 These are available in 2.6 and later kernels only. They indicate that
1224 autodetected MD arrays should be created as partitionable arrays, with
1225 a different major device number to the original non-partitionable md
1226 arrays. The device number is listed as
1232 .B md_mod.start_ro=1
1234 .B /sys/module/md_mod/parameters/start_ro
1235 This tells md to start all arrays in read-only mode. This is a soft
1236 read-only that will automatically switch to read-write on the first
1237 write request. However until that write request, nothing is written
1238 to any device by md, and in particular, no resync or recovery
1239 operation is started.
1242 .B md_mod.start_dirty_degraded=1
1244 .B /sys/module/md_mod/parameters/start_dirty_degraded
1245 As mentioned above, md will not normally start a RAID4, RAID5, or
1246 RAID6 that is both dirty and degraded as this situation can imply
1247 hidden data loss. This can be awkward if the root filesystem is
1248 affected. Using this module parameter allows such arrays to be started
1249 at boot time. It should be understood that there is a real (though
1250 small) risk of data corruption in this situation.
1253 .BI md= n , dev , dev ,...
1255 .BI md=d n , dev , dev ,...
1256 This tells the md driver to assemble
1258 from the listed devices. It is only necessary to start the device
1259 holding the root filesystem this way. Other arrays are best started
1260 once the system is booted.
1264 immediately after the
1266 indicates that a partitionable device (e.g.
1268 should be created rather than the original non-partitionable device.
1271 .BI md= n , l , c , i , dev...
1272 This tells the md driver to assemble a legacy RAID0 or LINEAR array
1273 without a superblock.
1275 gives the md device number,
1277 gives the level, 0 for RAID0 or \-1 for LINEAR,
1279 gives the chunk size as a base-2 logarithm offset by twelve, so 0
1280 means 4K, 1 means 8K.
1282 is ignored (legacy support).
1287 Contains information about the status of currently running array.
1289 .B /proc/sys/dev/raid/speed_limit_min
1290 A readable and writable file that reflects the current "goal" rebuild
1291 speed for times when non-rebuild activity is current on an array.
1292 The speed is in Kibibytes per second, and is a per-device rate, not a
1293 per-array rate (which means that an array with more disks will shuffle
1294 more data for a given speed). The default is 1000.
1297 .B /proc/sys/dev/raid/speed_limit_max
1298 A readable and writable file that reflects the current "goal" rebuild
1299 speed for times when no non-rebuild activity is current on an array.
1300 The default is 200,000.