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.
199 A RAID1 array is also known as a mirrored set (though mirrors tend to
200 provide reflected images, which RAID1 does not) or a plex.
202 Once initialised, each device in a RAID1 array contains exactly the
203 same data. Changes are written to all devices in parallel. Data is
204 read from any one device. The driver attempts to distribute read
205 requests across all devices to maximise performance.
207 All devices in a RAID1 array should be the same size. If they are
208 not, then only the amount of space available on the smallest device is
209 used (any extra space on other devices is wasted).
211 Note that the read balancing done by the driver does not make the RAID1
212 performance profile be the same as for RAID0; a single stream of
213 sequential input will not be accelerated (e.g. a single dd), but
214 multiple sequential streams or a random workload will use more than one
215 spindle. In theory, having an N-disk RAID1 will allow N sequential
216 threads to read from all disks.
218 Individual devices in a RAID1 can be marked as "write-mostly".
219 These drives are excluded from the normal read balancing and will only
220 be read from when there is no other option. This can be useful for
221 devices connected over a slow link.
225 A RAID4 array is like a RAID0 array with an extra device for storing
226 parity. This device is the last of the active devices in the
227 array. Unlike RAID0, RAID4 also requires that all stripes span all
228 drives, so extra space on devices that are larger than the smallest is
231 When any block in a RAID4 array is modified, the parity block for that
232 stripe (i.e. the block in the parity device at the same device offset
233 as the stripe) is also modified so that the parity block always
234 contains the "parity" for the whole stripe. I.e. its content is
235 equivalent to the result of performing an exclusive-or operation
236 between all the data blocks in the stripe.
238 This allows the array to continue to function if one device fails.
239 The data that was on that device can be calculated as needed from the
240 parity block and the other data blocks.
244 RAID5 is very similar to RAID4. The difference is that the parity
245 blocks for each stripe, instead of being on a single device, are
246 distributed across all devices. This allows more parallelism when
247 writing, as two different block updates will quite possibly affect
248 parity blocks on different devices so there is less contention.
250 This also allows more parallelism when reading, as read requests are
251 distributed over all the devices in the array instead of all but one.
255 RAID6 is similar to RAID5, but can handle the loss of any \fItwo\fP
256 devices without data loss. Accordingly, it requires N+2 drives to
257 store N drives worth of data.
259 The performance for RAID6 is slightly lower but comparable to RAID5 in
260 normal mode and single disk failure mode. It is very slow in dual
261 disk failure mode, however.
265 RAID10 provides a combination of RAID1 and RAID0, and is sometimes known
266 as RAID1+0. Every datablock is duplicated some number of times, and
267 the resulting collection of datablocks are distributed over multiple
270 When configuring a RAID10 array, it is necessary to specify the number
271 of replicas of each data block that are required (this will usually
272 be\ 2) and whether their layout should be "near", "far" or "offset"
273 (with "offset" being available since Linux\ 2.6.18).
275 .B About the RAID10 Layout Examples:
277 The examples below visualise the chunk distribution on the underlying
278 devices for the respective layout.
280 For simplicity it is assumed that the size of the chunks equals the
281 size of the blocks of the underlying devices as well as those of the
282 RAID10 device exported by the kernel (for example \fB/dev/md/\fPname).
284 Therefore the chunks\ /\ chunk numbers map directly to the blocks\ /\
285 block addresses of the exported RAID10 device.
287 Decimal numbers (0,\ 1, 2,\ ...) are the chunks of the RAID10 and due
288 to the above assumption also the blocks and block addresses of the
289 exported RAID10 device.
291 Repeated numbers mean copies of a chunk\ /\ block (obviously on
292 different underlying devices).
294 Hexadecimal numbers (0x00,\ 0x01, 0x02,\ ...) are the block addresses
295 of the underlying devices.
299 When "near" replicas are chosen, the multiple copies of a given chunk are laid
300 out consecutively ("as close to each other as possible") across the stripes of
303 With an even number of devices, they will likely (unless some misalignment is
304 present) lay at the very same offset on the different devices.
306 This is as the "classic" RAID1+0; that is two groups of mirrored devices (in the
307 example below the groups Device\ #1\ /\ #2 and Device\ #3\ /\ #4 are each a
308 RAID1) both in turn forming a striped RAID0.
311 .B Example with 2\ copies per chunk and an even number\ (4) of devices:
316 | - | - | - | - | - |
317 | C | C | C | C | C |
318 | C | C | C | C | C |
319 | C | C | C | C | C |
320 | C | C | C | C | C |
321 | C | C | C | C | C |
322 | C | C | C | C | C |
323 | - | - | - | - | - |
329 ;Device #1;Device #2;Device #3;Device #4
332 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.
334 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.
336 ;\\---------v---------/;\\---------v---------/
338 ;\\---------------------v---------------------/
343 .B Example with 2\ copies per chunk and an odd number\ (5) of devices:
347 C | C | C | C | C | C |
348 | - | - | - | - | - | - |
349 | C | C | C | C | C | C |
350 | C | C | C | C | C | C |
351 | C | C | C | C | C | C |
352 | C | C | C | C | C | C |
353 | C | C | C | C | C | C |
354 | C | C | C | C | C | C |
355 | - | - | - | - | - | - |
358 ;Dev #1;Dev #2;Dev #3;Dev #4;Dev #5
361 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.
363 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.
364 0x80;317;318;318;319;319
370 When "far" replicas are chosen, the multiple copies of a given chunk
371 are laid out quite distant ("as far as reasonably possible") from each
374 First a complete sequence of all data blocks (that is all the data one
375 sees on the exported RAID10 block device) is striped over the
376 devices. Then another (though "shifted") complete sequence of all data
377 blocks; and so on (in the case of more than 2\ copies per chunk).
379 The "shift" needed to prevent placing copies of the same chunks on the
380 same devices is actually a cyclic permutation with offset\ 1 of each
381 of the stripes within a complete sequence of chunks.
383 The offset\ 1 is relative to the previous complete sequence of chunks,
384 so in case of more than 2\ copies per chunk one gets the following
387 1.\ complete sequence of chunks: offset\ =\ \ 0
389 2.\ complete sequence of chunks: offset\ =\ \ 1
391 3.\ complete sequence of chunks: offset\ =\ \ 2
395 n.\ complete sequence of chunks: offset\ =\ n-1
398 .B Example with 2\ copies per chunk and an even number\ (4) of devices:
403 | - | - | - | - | - |
404 | C | C | C | C | C | L
405 | C | C | C | C | C | L
406 | C | C | C | C | C | L
407 | C | C | C | C | C | L
408 | C | C | C | C | C | L
409 | C | C | C | C | C | L
410 | C | C | C | C | C | L
411 | C | C | C | C | C | L
412 | C | C | C | C | C | L
413 | C | C | C | C | C | L
414 | C | C | C | C | C | L
415 | C | C | C | C | C | L
416 | - | - | - | - | - |
419 ;Device #1;Device #2;Device #3;Device #4
423 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
425 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
426 0x40;252;253;254;255;/
429 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
431 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
432 0x80;255;252;253;254;/
437 .B Example with 2\ copies per chunk and an odd number\ (5) of devices:
441 C | C | C | C | C | C |
442 | - | - | - | - | - | - |
443 | C | C | C | C | C | C | L
444 | C | C | C | C | C | C | L
445 | C | C | C | C | C | C | L
446 | C | C | C | C | C | C | L
447 | C | C | C | C | C | C | L
448 | C | C | C | C | C | C | L
449 | C | C | C | C | C | C | L
450 | C | C | C | C | C | C | L
451 | C | C | C | C | C | C | L
452 | C | C | C | C | C | C | L
453 | C | C | C | C | C | C | L
454 | C | C | C | C | C | C | L
455 | - | - | - | - | - | - |
458 ;Dev #1;Dev #2;Dev #3;Dev #4;Dev #5
462 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
464 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
465 0x40;315;316;317;318;319;/
467 0x42;9;5;6;7;8;> [#]~
468 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
470 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;:
471 0x80;319;315;316;317;318;/
475 With [#]\ being the complete sequence of chunks and [#]~\ the cyclic permutation
476 with offset\ 1 thereof (in the case of more than 2 copies per chunk there would
477 be ([#]~)~,\ (([#]~)~)~,\ ...).
479 The advantage of this layout is that MD can easily spread sequential reads over
480 the devices, making them similar to RAID0 in terms of speed.
482 The cost is more seeking for writes, making them substantially slower.
485 \fB"offset" Layout\fP
486 When "offset" replicas are chosen, all the copies of a given chunk are
487 striped consecutively ("offset by the stripe length after each other")
490 Explained in detail, <number of devices> consecutive chunks are
491 striped over the devices, immediately followed by a "shifted" copy of
492 these chunks (and by further such "shifted" copies in the case of more
493 than 2\ copies per chunk).
495 This pattern repeats for all further consecutive chunks of the
496 exported RAID10 device (in other words: all further data blocks).
498 The "shift" needed to prevent placing copies of the same chunks on the
499 same devices is actually a cyclic permutation with offset\ 1 of each
500 of the striped copies of <number of devices> consecutive chunks.
502 The offset\ 1 is relative to the previous striped copy of <number of
503 devices> consecutive chunks, so in case of more than 2\ copies per
504 chunk one gets the following offsets:
506 1.\ <number of devices> consecutive chunks: offset\ =\ \ 0
508 2.\ <number of devices> consecutive chunks: offset\ =\ \ 1
510 3.\ <number of devices> consecutive chunks: offset\ =\ \ 2
514 n.\ <number of devices> consecutive chunks: offset\ =\ n-1
517 .B Example with 2\ copies per chunk and an even number\ (4) of devices:
522 | - | - | - | - | - |
523 | C | C | C | C | C | L
524 | C | C | C | C | C | L
525 | C | C | C | C | C | L
526 | C | C | C | C | C | L
527 | C | C | C | C | C | L
528 | C | C | C | C | C | L
529 | C | C | C | C | C | L
530 | C | C | C | C | C | L
531 | C | C | C | C | C | L
532 | - | - | - | - | - |
535 ;Device #1;Device #2;Device #3;Device #4
541 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;) \.\.\.
543 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;) \.\.\.
544 0x79;251;252;253;254;) EX
545 0x80;254;251;252;253;) EX~
550 .B Example with 2\ copies per chunk and an odd number\ (5) of devices:
554 C | C | C | C | C | C |
555 | - | - | - | - | - | - |
556 | C | C | C | C | C | C | L
557 | C | C | C | C | C | C | L
558 | C | C | C | C | C | C | L
559 | C | C | C | C | C | C | L
560 | C | C | C | C | C | C | L
561 | C | C | C | C | C | C | L
562 | C | C | C | C | C | C | L
563 | C | C | C | C | C | C | L
564 | C | C | C | C | C | C | L
565 | - | - | - | - | - | - |
568 ;Dev #1;Dev #2;Dev #3;Dev #4;Dev #5
574 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;) \.\.\.
576 \.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;\.\.\.;) \.\.\.
577 0x79;314;315;316;317;318;) EX
578 0x80;318;314;315;316;317;) EX~
582 With AA,\ AB,\ ..., AZ,\ BA,\ ... being the sets of <number of devices> consecutive
583 chunks and AA~,\ AB~,\ ..., AZ~,\ BA~,\ ... the cyclic permutations with offset\ 1
584 thereof (in the case of more than 2 copies per chunk there would be (AA~)~,\ ...
585 as well as ((AA~)~)~,\ ... and so on).
587 This should give similar read characteristics to "far" if a suitably large chunk
588 size is used, but without as much seeking for writes.
592 It should be noted that the number of devices in a RAID10 array need
593 not be a multiple of the number of replica of each data block; however,
594 there must be at least as many devices as replicas.
596 If, for example, an array is created with 5 devices and 2 replicas,
597 then space equivalent to 2.5 of the devices will be available, and
598 every block will be stored on two different devices.
600 Finally, it is possible to have an array with both "near" and "far"
601 copies. If an array is configured with 2 near copies and 2 far
602 copies, then there will be a total of 4 copies of each block, each on
603 a different drive. This is an artifact of the implementation and is
604 unlikely to be of real value.
608 MULTIPATH is not really a RAID at all as there is only one real device
609 in a MULTIPATH md array. However there are multiple access points
610 (paths) to this device, and one of these paths might fail, so there
611 are some similarities.
613 A MULTIPATH array is composed of a number of logically different
614 devices, often fibre channel interfaces, that all refer the the same
615 real device. If one of these interfaces fails (e.g. due to cable
616 problems), the MULTIPATH driver will attempt to redirect requests to
619 The MULTIPATH drive is not receiving any ongoing development and
620 should be considered a legacy driver. The device-mapper based
621 multipath drivers should be preferred for new installations.
624 The FAULTY md module is provided for testing purposes. A FAULTY array
625 has exactly one component device and is normally assembled without a
626 superblock, so the md array created provides direct access to all of
627 the data in the component device.
629 The FAULTY module may be requested to simulate faults to allow testing
630 of other md levels or of filesystems. Faults can be chosen to trigger
631 on read requests or write requests, and can be transient (a subsequent
632 read/write at the address will probably succeed) or persistent
633 (subsequent read/write of the same address will fail). Further, read
634 faults can be "fixable" meaning that they persist until a write
635 request at the same address.
637 Fault types can be requested with a period. In this case, the fault
638 will recur repeatedly after the given number of requests of the
639 relevant type. For example if persistent read faults have a period of
640 100, then every 100th read request would generate a fault, and the
641 faulty sector would be recorded so that subsequent reads on that
642 sector would also fail.
644 There is a limit to the number of faulty sectors that are remembered.
645 Faults generated after this limit is exhausted are treated as
648 The list of faulty sectors can be flushed, and the active list of
649 failure modes can be cleared.
653 When changes are made to a RAID1, RAID4, RAID5, RAID6, or RAID10 array
654 there is a possibility of inconsistency for short periods of time as
655 each update requires at least two block to be written to different
656 devices, and these writes probably won't happen at exactly the same
657 time. Thus if a system with one of these arrays is shutdown in the
658 middle of a write operation (e.g. due to power failure), the array may
661 To handle this situation, the md driver marks an array as "dirty"
662 before writing any data to it, and marks it as "clean" when the array
663 is being disabled, e.g. at shutdown. If the md driver finds an array
664 to be dirty at startup, it proceeds to correct any possibly
665 inconsistency. For RAID1, this involves copying the contents of the
666 first drive onto all other drives. For RAID4, RAID5 and RAID6 this
667 involves recalculating the parity for each stripe and making sure that
668 the parity block has the correct data. For RAID10 it involves copying
669 one of the replicas of each block onto all the others. This process,
670 known as "resynchronising" or "resync" is performed in the background.
671 The array can still be used, though possibly with reduced performance.
673 If a RAID4, RAID5 or RAID6 array is degraded (missing at least one
674 drive, two for RAID6) when it is restarted after an unclean shutdown, it cannot
675 recalculate parity, and so it is possible that data might be
676 undetectably corrupted. The 2.4 md driver
678 alert the operator to this condition. The 2.6 md driver will fail to
679 start an array in this condition without manual intervention, though
680 this behaviour can be overridden by a kernel parameter.
684 If the md driver detects a write error on a device in a RAID1, RAID4,
685 RAID5, RAID6, or RAID10 array, it immediately disables that device
686 (marking it as faulty) and continues operation on the remaining
687 devices. If there are spare drives, the driver will start recreating
688 on one of the spare drives the data which was on that failed drive,
689 either by copying a working drive in a RAID1 configuration, or by
690 doing calculations with the parity block on RAID4, RAID5 or RAID6, or
691 by finding and copying originals for RAID10.
693 In kernels prior to about 2.6.15, a read error would cause the same
694 effect as a write error. In later kernels, a read-error will instead
695 cause md to attempt a recovery by overwriting the bad block. i.e. it
696 will find the correct data from elsewhere, write it over the block
697 that failed, and then try to read it back again. If either the write
698 or the re-read fail, md will treat the error the same way that a write
699 error is treated, and will fail the whole device.
701 While this recovery process is happening, the md driver will monitor
702 accesses to the array and will slow down the rate of recovery if other
703 activity is happening, so that normal access to the array will not be
704 unduly affected. When no other activity is happening, the recovery
705 process proceeds at full speed. The actual speed targets for the two
706 different situations can be controlled by the
710 control files mentioned below.
712 .SS SCRUBBING AND MISMATCHES
714 As storage devices can develop bad blocks at any time it is valuable
715 to regularly read all blocks on all devices in an array so as to catch
716 such bad blocks early. This process is called
719 md arrays can be scrubbed by writing either
727 directory for the device.
729 Requesting a scrub will cause
731 to read every block on every device in the array, and check that the
732 data is consistent. For RAID1 and RAID10, this means checking that the copies
733 are identical. For RAID4, RAID5, RAID6 this means checking that the
734 parity block is (or blocks are) correct.
736 If a read error is detected during this process, the normal read-error
737 handling causes correct data to be found from other devices and to be
738 written back to the faulty device. In many case this will
743 If all blocks read successfully but are found to not be consistent,
744 then this is regarded as a
749 was used, then no action is taken to handle the mismatch, it is simply
753 was used, then a mismatch will be repaired in the same way that
755 repairs arrays. For RAID5/RAID6 new parity blocks are written. For RAID1/RAID10,
756 all but one block are overwritten with the content of that one block.
758 A count of mismatches is recorded in the
761 .IR md/mismatch_cnt .
762 This is set to zero when a
763 scrub starts and is incremented whenever a sector is
764 found that is a mismatch.
766 normally works in units much larger than a single sector and when it
767 finds a mismatch, it does not determine exactly how many actual sectors were
768 affected but simply adds the number of sectors in the IO unit that was
769 used. So a value of 128 could simply mean that a single 64KB check
770 found an error (128 x 512bytes = 64KB).
772 If an array is created by
776 then a subsequent check could be expected to find some mismatches.
778 On a truly clean RAID5 or RAID6 array, any mismatches should indicate
779 a hardware problem at some level - software issues should never cause
782 However on RAID1 and RAID10 it is possible for software issues to
783 cause a mismatch to be reported. This does not necessarily mean that
784 the data on the array is corrupted. It could simply be that the
785 system does not care what is stored on that part of the array - it is
788 The most likely cause for an unexpected mismatch on RAID1 or RAID10
789 occurs if a swap partition or swap file is stored on the array.
791 When the swap subsystem wants to write a page of memory out, it flags
792 the page as 'clean' in the memory manager and requests the swap device
793 to write it out. It is quite possible that the memory will be
794 changed while the write-out is happening. In that case the 'clean'
795 flag will be found to be clear when the write completes and so the
796 swap subsystem will simply forget that the swapout had been attempted,
797 and will possibly choose a different page to write out.
799 If the swap device was on RAID1 (or RAID10), then the data is sent
800 from memory to a device twice (or more depending on the number of
801 devices in the array). Thus it is possible that the memory gets changed
802 between the times it is sent, so different data can be written to
803 the different devices in the array. This will be detected by
805 as a mismatch. However it does not reflect any corruption as the
806 block where this mismatch occurs is being treated by the swap system as
807 being empty, and the data will never be read from that block.
809 It is conceivable for a similar situation to occur on non-swap files,
810 though it is less likely.
814 value can not be interpreted very reliably on RAID1 or RAID10,
815 especially when the device is used for swap.
818 .SS BITMAP WRITE-INTENT LOGGING
822 supports a bitmap based write-intent log. If configured, the bitmap
823 is used to record which blocks of the array may be out of sync.
824 Before any write request is honoured, md will make sure that the
825 corresponding bit in the log is set. After a period of time with no
826 writes to an area of the array, the corresponding bit will be cleared.
828 This bitmap is used for two optimisations.
830 Firstly, after an unclean shutdown, the resync process will consult
831 the bitmap and only resync those blocks that correspond to bits in the
832 bitmap that are set. This can dramatically reduce resync time.
834 Secondly, when a drive fails and is removed from the array, md stops
835 clearing bits in the intent log. If that same drive is re-added to
836 the array, md will notice and will only recover the sections of the
837 drive that are covered by bits in the intent log that are set. This
838 can allow a device to be temporarily removed and reinserted without
839 causing an enormous recovery cost.
841 The intent log can be stored in a file on a separate device, or it can
842 be stored near the superblocks of an array which has superblocks.
844 It is possible to add an intent log to an active array, or remove an
845 intent log if one is present.
847 In 2.6.13, intent bitmaps are only supported with RAID1. Other levels
848 with redundancy are supported from 2.6.15.
852 From Linux 3.5 each device in an
854 array can store a list of known-bad-blocks. This list is 4K in size
855 and usually positioned at the end of the space between the superblock
858 When a block cannot be read and cannot be repaired by writing data
859 recovered from other devices, the address of the block is stored in
860 the bad block list. Similarly if an attempt to write a block fails,
861 the address will be recorded as a bad block. If attempting to record
862 the bad block fails, the whole device will be marked faulty.
864 Attempting to read from a known bad block will cause a read error.
865 Attempting to write to a known bad block will be ignored if any write
866 errors have been reported by the device. If there have been no write
867 errors then the data will be written to the known bad block and if
868 that succeeds, the address will be removed from the list.
870 This allows an array to fail more gracefully - a few blocks on
871 different devices can be faulty without taking the whole array out of
874 The list is particularly useful when recovering to a spare. If a few blocks
875 cannot be read from the other devices, the bulk of the recovery can
876 complete and those few bad blocks will be recorded in the bad block list.
878 .SS RAID456 WRITE JOURNAL
880 Due to non-atomicity nature of RAID write operations, interruption of
881 write operations (system crash, etc.) to RAID456 array can lead to
882 inconsistent parity and data loss (so called RAID-5 write hole).
884 To plug the write hole, from Linux 4.4 (to be confirmed),
886 supports write ahead journal for RAID456. When the array is created,
887 an additional journal device can be added to the array through
889 option. The RAID write journal works similar to file system journals.
890 Before writing to the data disks, md persists data AND parity of the
891 stripe to the journal device. After crashes, md searches the journal
892 device for incomplete write operations, and replay them to the data
895 When the journal device fails, the RAID array is forced to run in
902 supports WRITE-BEHIND on RAID1 arrays.
904 This allows certain devices in the array to be flagged as
906 MD will only read from such devices if there is no
909 If a write-intent bitmap is also provided, write requests to
910 write-mostly devices will be treated as write-behind requests and md
911 will not wait for writes to those requests to complete before
912 reporting the write as complete to the filesystem.
914 This allows for a RAID1 with WRITE-BEHIND to be used to mirror data
915 over a slow link to a remote computer (providing the link isn't too
916 slow). The extra latency of the remote link will not slow down normal
917 operations, but the remote system will still have a reasonably
918 up-to-date copy of all data.
925 supports FAILFAST for RAID1 and RAID10 arrays. This is a flag that
926 can be set on individual drives, though it is usually set on all
927 drives, or no drives.
931 sends an I/O request to a drive that is marked as FAILFAST, and when
932 the array could survive the loss of that drive without losing data,
934 will request that the underlying device does not perform any retries.
935 This means that a failure will be reported to
937 promptly, and it can mark the device as faulty and continue using the
940 cannot control the timeout that the underlying devices use to
941 determine failure. Any changes desired to that timeout must be set
942 explictly on the underlying device, separately from using
945 If a FAILFAST request does fail, and if it is still safe to mark the
946 device as faulty without data loss, that will be done and the array
947 will continue functioning on a reduced number of devices. If it is not
948 possible to safely mark the device as faulty,
950 will retry the request without disabling retries in the underlying
953 will not attempt to repair read errors on a device marked as FAILFAST
954 by writing out the correct. It will just mark the device as faulty.
956 FAILFAST is appropriate for storage arrays that have a low probability
957 of true failure, but will sometimes introduce unacceptable delays to
958 I/O requests while performing internal maintenance. The value of
959 setting FAILFAST involves a trade-off. The gain is that the chance of
960 unacceptable delays is substantially reduced. The cost is that the
961 unlikely event of data-loss on one device is slightly more likely to
962 result in data-loss for the array.
964 When a device in an array using FAILFAST is marked as faulty, it will
965 usually become usable again in a short while.
967 makes no attempt to detect that possibility. Some separate
968 mechanism, tuned to the specific details of the expected failure modes,
969 needs to be created to monitor devices to see when they return to full
970 functionality, and to then re-add them to the array. In order of
971 this "re-add" functionality to be effective, an array using FAILFAST
972 should always have a write-intent bitmap.
979 is the processes of re-arranging the data stored in each stripe into a
980 new layout. This might involve changing the number of devices in the
981 array (so the stripes are wider), changing the chunk size (so stripes
982 are deeper or shallower), or changing the arrangement of data and
983 parity (possibly changing the RAID level, e.g. 1 to 5 or 5 to 6).
985 As of Linux 2.6.35, md can reshape a RAID4, RAID5, or RAID6 array to
986 have a different number of devices (more or fewer) and to have a
987 different layout or chunk size. It can also convert between these
988 different RAID levels. It can also convert between RAID0 and RAID10,
989 and between RAID0 and RAID4 or RAID5.
990 Other possibilities may follow in future kernels.
992 During any stripe process there is a 'critical section' during which
993 live data is being overwritten on disk. For the operation of
994 increasing the number of drives in a RAID5, this critical section
995 covers the first few stripes (the number being the product of the old
996 and new number of devices). After this critical section is passed,
997 data is only written to areas of the array which no longer hold live
998 data \(em the live data has already been located away.
1000 For a reshape which reduces the number of devices, the 'critical
1001 section' is at the end of the reshape process.
1003 md is not able to ensure data preservation if there is a crash
1004 (e.g. power failure) during the critical section. If md is asked to
1005 start an array which failed during a critical section of restriping,
1006 it will fail to start the array.
1008 To deal with this possibility, a user-space program must
1010 Disable writes to that section of the array (using the
1014 take a copy of the data somewhere (i.e. make a backup),
1016 allow the process to continue and invalidate the backup and restore
1017 write access once the critical section is passed, and
1019 provide for restoring the critical data before restarting the array
1020 after a system crash.
1024 versions from 2.4 do this for growing a RAID5 array.
1026 For operations that do not change the size of the array, like simply
1027 increasing chunk size, or converting RAID5 to RAID6 with one extra
1028 device, the entire process is the critical section. In this case, the
1029 restripe will need to progress in stages, as a section is suspended,
1030 backed up, restriped, and released.
1033 Each block device appears as a directory in
1035 (which is usually mounted at
1037 For MD devices, this directory will contain a subdirectory called
1039 which contains various files for providing access to information about
1042 This interface is documented more fully in the file
1043 .B Documentation/md.txt
1044 which is distributed with the kernel sources. That file should be
1045 consulted for full documentation. The following are just a selection
1046 of attribute files that are available.
1049 .B md/sync_speed_min
1050 This value, if set, overrides the system-wide setting in
1051 .B /proc/sys/dev/raid/speed_limit_min
1052 for this array only.
1055 to this file will cause the system-wide setting to have effect.
1058 .B md/sync_speed_max
1059 This is the partner of
1060 .B md/sync_speed_min
1062 .B /proc/sys/dev/raid/speed_limit_max
1067 This can be used to monitor and control the resync/recovery process of
1069 In particular, writing "check" here will cause the array to read all
1070 data block and check that they are consistent (e.g. parity is correct,
1071 or all mirror replicas are the same). Any discrepancies found are
1075 A count of problems found will be stored in
1076 .BR md/mismatch_count .
1078 Alternately, "repair" can be written which will cause the same check
1079 to be performed, but any errors will be corrected.
1081 Finally, "idle" can be written to stop the check/repair process.
1084 .B md/stripe_cache_size
1085 This is only available on RAID5 and RAID6. It records the size (in
1086 pages per device) of the stripe cache which is used for synchronising
1087 all write operations to the array and all read operations if the array
1088 is degraded. The default is 256. Valid values are 17 to 32768.
1089 Increasing this number can increase performance in some situations, at
1090 some cost in system memory. Note, setting this value too high can
1091 result in an "out of memory" condition for the system.
1093 memory_consumed = system_page_size * nr_disks * stripe_cache_size
1096 .B md/preread_bypass_threshold
1097 This is only available on RAID5 and RAID6. This variable sets the
1098 number of times MD will service a full-stripe-write before servicing a
1099 stripe that requires some "prereading". For fairness this defaults to
1100 1. Valid values are 0 to stripe_cache_size. Setting this to 0
1101 maximizes sequential-write throughput at the cost of fairness to threads
1102 doing small or random writes.
1104 .SS KERNEL PARAMETERS
1106 The md driver recognised several different kernel parameters.
1108 .B raid=noautodetect
1109 This will disable the normal detection of md arrays that happens at
1110 boot time. If a drive is partitioned with MS-DOS style partitions,
1111 then if any of the 4 main partitions has a partition type of 0xFD,
1112 then that partition will normally be inspected to see if it is part of
1113 an MD array, and if any full arrays are found, they are started. This
1114 kernel parameter disables this behaviour.
1117 .B raid=partitionable
1120 These are available in 2.6 and later kernels only. They indicate that
1121 autodetected MD arrays should be created as partitionable arrays, with
1122 a different major device number to the original non-partitionable md
1123 arrays. The device number is listed as
1129 .B md_mod.start_ro=1
1131 .B /sys/module/md_mod/parameters/start_ro
1132 This tells md to start all arrays in read-only mode. This is a soft
1133 read-only that will automatically switch to read-write on the first
1134 write request. However until that write request, nothing is written
1135 to any device by md, and in particular, no resync or recovery
1136 operation is started.
1139 .B md_mod.start_dirty_degraded=1
1141 .B /sys/module/md_mod/parameters/start_dirty_degraded
1142 As mentioned above, md will not normally start a RAID4, RAID5, or
1143 RAID6 that is both dirty and degraded as this situation can imply
1144 hidden data loss. This can be awkward if the root filesystem is
1145 affected. Using this module parameter allows such arrays to be started
1146 at boot time. It should be understood that there is a real (though
1147 small) risk of data corruption in this situation.
1150 .BI md= n , dev , dev ,...
1152 .BI md=d n , dev , dev ,...
1153 This tells the md driver to assemble
1155 from the listed devices. It is only necessary to start the device
1156 holding the root filesystem this way. Other arrays are best started
1157 once the system is booted.
1161 immediately after the
1163 indicates that a partitionable device (e.g.
1165 should be created rather than the original non-partitionable device.
1168 .BI md= n , l , c , i , dev...
1169 This tells the md driver to assemble a legacy RAID0 or LINEAR array
1170 without a superblock.
1172 gives the md device number,
1174 gives the level, 0 for RAID0 or \-1 for LINEAR,
1176 gives the chunk size as a base-2 logarithm offset by twelve, so 0
1177 means 4K, 1 means 8K.
1179 is ignored (legacy support).
1184 Contains information about the status of currently running array.
1186 .B /proc/sys/dev/raid/speed_limit_min
1187 A readable and writable file that reflects the current "goal" rebuild
1188 speed for times when non-rebuild activity is current on an array.
1189 The speed is in Kibibytes per second, and is a per-device rate, not a
1190 per-array rate (which means that an array with more disks will shuffle
1191 more data for a given speed). The default is 1000.
1194 .B /proc/sys/dev/raid/speed_limit_max
1195 A readable and writable file that reflects the current "goal" rebuild
1196 speed for times when no non-rebuild activity is current on an array.
1197 The default is 200,000.