766 lines · plain
1RAID arrays2===========3 4Boot time assembly of RAID arrays5---------------------------------6 7Tools that manage md devices can be found at8 https://www.kernel.org/pub/linux/utils/raid/9 10 11You can boot with your md device with the following kernel command12lines:13 14for old raid arrays without persistent superblocks::15 16 md=<md device no.>,<raid level>,<chunk size factor>,<fault level>,dev0,dev1,...,devn17 18for raid arrays with persistent superblocks::19 20 md=<md device no.>,dev0,dev1,...,devn21 22or, to assemble a partitionable array::23 24 md=d<md device no.>,dev0,dev1,...,devn25 26``md device no.``27+++++++++++++++++28 29The number of the md device30 31================= =========32``md device no.`` device33================= =========34 0 md035 1 md136 2 md237 3 md338 4 md439================= =========40 41``raid level``42++++++++++++++43 44level of the RAID array45 46=============== =============47``raid level`` level48=============== =============49-1 linear mode500 striped mode51=============== =============52 53other modes are only supported with persistent super blocks54 55``chunk size factor``56+++++++++++++++++++++57 58(raid-0 and raid-1 only)59 60Set the chunk size as 4k << n.61 62``fault level``63+++++++++++++++64 65Totally ignored66 67``dev0`` to ``devn``68++++++++++++++++++++69 70e.g. ``/dev/hda1``, ``/dev/hdc1``, ``/dev/sda1``, ``/dev/sdb1``71 72A possible loadlin line (Harald Hoyer <HarryH@Royal.Net>) looks like this::73 74 e:\loadlin\loadlin e:\zimage root=/dev/md0 md=0,0,4,0,/dev/hdb2,/dev/hdc3 ro75 76 77Boot time autodetection of RAID arrays78--------------------------------------79 80When md is compiled into the kernel (not as module), partitions of81type 0xfd are scanned and automatically assembled into RAID arrays.82This autodetection may be suppressed with the kernel parameter83``raid=noautodetect``. As of kernel 2.6.9, only drives with a type 084superblock can be autodetected and run at boot time.85 86The kernel parameter ``raid=partitionable`` (or ``raid=part``) means87that all auto-detected arrays are assembled as partitionable.88 89Boot time assembly of degraded/dirty arrays90-------------------------------------------91 92If a raid5 or raid6 array is both dirty and degraded, it could have93undetectable data corruption. This is because the fact that it is94``dirty`` means that the parity cannot be trusted, and the fact that it95is degraded means that some datablocks are missing and cannot reliably96be reconstructed (due to no parity).97 98For this reason, md will normally refuse to start such an array. This99requires the sysadmin to take action to explicitly start the array100despite possible corruption. This is normally done with::101 102 mdadm --assemble --force ....103 104This option is not really available if the array has the root105filesystem on it. In order to support this booting from such an106array, md supports a module parameter ``start_dirty_degraded`` which,107when set to 1, bypassed the checks and will allows dirty degraded108arrays to be started.109 110So, to boot with a root filesystem of a dirty degraded raid 5 or 6, use::111 112 md-mod.start_dirty_degraded=1113 114 115Superblock formats116------------------117 118The md driver can support a variety of different superblock formats.119Currently, it supports superblock formats ``0.90.0`` and the ``md-1`` format120introduced in the 2.5 development series.121 122The kernel will autodetect which format superblock is being used.123 124Superblock format ``0`` is treated differently to others for legacy125reasons - it is the original superblock format.126 127 128General Rules - apply for all superblock formats129------------------------------------------------130 131An array is ``created`` by writing appropriate superblocks to all132devices.133 134It is ``assembled`` by associating each of these devices with an135particular md virtual device. Once it is completely assembled, it can136be accessed.137 138An array should be created by a user-space tool. This will write139superblocks to all devices. It will usually mark the array as140``unclean``, or with some devices missing so that the kernel md driver141can create appropriate redundancy (copying in raid 1, parity142calculation in raid 4/5).143 144When an array is assembled, it is first initialized with the145SET_ARRAY_INFO ioctl. This contains, in particular, a major and minor146version number. The major version number selects which superblock147format is to be used. The minor number might be used to tune handling148of the format, such as suggesting where on each device to look for the149superblock.150 151Then each device is added using the ADD_NEW_DISK ioctl. This152provides, in particular, a major and minor number identifying the153device to add.154 155The array is started with the RUN_ARRAY ioctl.156 157Once started, new devices can be added. They should have an158appropriate superblock written to them, and then be passed in with159ADD_NEW_DISK.160 161Devices that have failed or are not yet active can be detached from an162array using HOT_REMOVE_DISK.163 164 165Specific Rules that apply to format-0 super block arrays, and arrays with no superblock (non-persistent)166--------------------------------------------------------------------------------------------------------167 168An array can be ``created`` by describing the array (level, chunksize169etc) in a SET_ARRAY_INFO ioctl. This must have ``major_version==0`` and170``raid_disks != 0``.171 172Then uninitialized devices can be added with ADD_NEW_DISK. The173structure passed to ADD_NEW_DISK must specify the state of the device174and its role in the array.175 176Once started with RUN_ARRAY, uninitialized spares can be added with177HOT_ADD_DISK.178 179 180MD devices in sysfs181-------------------182 183md devices appear in sysfs (``/sys``) as regular block devices,184e.g.::185 186 /sys/block/md0187 188Each ``md`` device will contain a subdirectory called ``md`` which189contains further md-specific information about the device.190 191All md devices contain:192 193 level194 a text file indicating the ``raid level``. e.g. raid0, raid1,195 raid5, linear, multipath, faulty.196 If no raid level has been set yet (array is still being197 assembled), the value will reflect whatever has been written198 to it, which may be a name like the above, or may be a number199 such as ``0``, ``5``, etc.200 201 raid_disks202 a text file with a simple number indicating the number of devices203 in a fully functional array. If this is not yet known, the file204 will be empty. If an array is being resized this will contain205 the new number of devices.206 Some raid levels allow this value to be set while the array is207 active. This will reconfigure the array. Otherwise it can only208 be set while assembling an array.209 A change to this attribute will not be permitted if it would210 reduce the size of the array. To reduce the number of drives211 in an e.g. raid5, the array size must first be reduced by212 setting the ``array_size`` attribute.213 214 chunk_size215 This is the size in bytes for ``chunks`` and is only relevant to216 raid levels that involve striping (0,4,5,6,10). The address space217 of the array is conceptually divided into chunks and consecutive218 chunks are striped onto neighbouring devices.219 The size should be at least PAGE_SIZE (4k) and should be a power220 of 2. This can only be set while assembling an array221 222 layout223 The ``layout`` for the array for the particular level. This is224 simply a number that is interpreted differently by different225 levels. It can be written while assembling an array.226 227 array_size228 This can be used to artificially constrain the available space in229 the array to be less than is actually available on the combined230 devices. Writing a number (in Kilobytes) which is less than231 the available size will set the size. Any reconfiguration of the232 array (e.g. adding devices) will not cause the size to change.233 Writing the word ``default`` will cause the effective size of the234 array to be whatever size is actually available based on235 ``level``, ``chunk_size`` and ``component_size``.236 237 This can be used to reduce the size of the array before reducing238 the number of devices in a raid4/5/6, or to support external239 metadata formats which mandate such clipping.240 241 reshape_position242 This is either ``none`` or a sector number within the devices of243 the array where ``reshape`` is up to. If this is set, the three244 attributes mentioned above (raid_disks, chunk_size, layout) can245 potentially have 2 values, an old and a new value. If these246 values differ, reading the attribute returns::247 248 new (old)249 250 and writing will effect the ``new`` value, leaving the ``old``251 unchanged.252 253 component_size254 For arrays with data redundancy (i.e. not raid0, linear, faulty,255 multipath), all components must be the same size - or at least256 there must a size that they all provide space for. This is a key257 part or the geometry of the array. It is measured in sectors258 and can be read from here. Writing to this value may resize259 the array if the personality supports it (raid1, raid5, raid6),260 and if the component drives are large enough.261 262 metadata_version263 This indicates the format that is being used to record metadata264 about the array. It can be 0.90 (traditional format), 1.0, 1.1,265 1.2 (newer format in varying locations) or ``none`` indicating that266 the kernel isn't managing metadata at all.267 Alternately it can be ``external:`` followed by a string which268 is set by user-space. This indicates that metadata is managed269 by a user-space program. Any device failure or other event that270 requires a metadata update will cause array activity to be271 suspended until the event is acknowledged.272 273 resync_start274 The point at which resync should start. If no resync is needed,275 this will be a very large number (or ``none`` since 2.6.30-rc1). At276 array creation it will default to 0, though starting the array as277 ``clean`` will set it much larger.278 279 new_dev280 This file can be written but not read. The value written should281 be a block device number as major:minor. e.g. 8:0282 This will cause that device to be attached to the array, if it is283 available. It will then appear at md/dev-XXX (depending on the284 name of the device) and further configuration is then possible.285 286 safe_mode_delay287 When an md array has seen no write requests for a certain period288 of time, it will be marked as ``clean``. When another write289 request arrives, the array is marked as ``dirty`` before the write290 commences. This is known as ``safe_mode``.291 The ``certain period`` is controlled by this file which stores the292 period as a number of seconds. The default is 200msec (0.200).293 Writing a value of 0 disables safemode.294 295 array_state296 This file contains a single word which describes the current297 state of the array. In many cases, the state can be set by298 writing the word for the desired state, however some states299 cannot be explicitly set, and some transitions are not allowed.300 301 Select/poll works on this file. All changes except between302 Active_idle and active (which can be frequent and are not303 very interesting) are notified. active->active_idle is304 reported if the metadata is externally managed.305 306 clear307 No devices, no size, no level308 309 Writing is equivalent to STOP_ARRAY ioctl310 311 inactive312 May have some settings, but array is not active313 all IO results in error314 315 When written, doesn't tear down array, but just stops it316 317 suspended (not supported yet)318 All IO requests will block. The array can be reconfigured.319 320 Writing this, if accepted, will block until array is quiescent321 322 readonly323 no resync can happen. no superblocks get written.324 325 Write requests fail326 327 read-auto328 like readonly, but behaves like ``clean`` on a write request.329 330 clean331 no pending writes, but otherwise active.332 333 When written to inactive array, starts without resync334 335 If a write request arrives then336 if metadata is known, mark ``dirty`` and switch to ``active``.337 if not known, block and switch to write-pending338 339 If written to an active array that has pending writes, then fails.340 active341 fully active: IO and resync can be happening.342 When written to inactive array, starts with resync343 344 write-pending345 clean, but writes are blocked waiting for ``active`` to be written.346 347 active-idle348 like active, but no writes have been seen for a while (safe_mode_delay).349 350 bitmap/location351 This indicates where the write-intent bitmap for the array is352 stored.353 354 It can be one of ``none``, ``file`` or ``[+-]N``.355 ``file`` may later be extended to ``file:/file/name``356 ``[+-]N`` means that many sectors from the start of the metadata.357 358 This is replicated on all devices. For arrays with externally359 managed metadata, the offset is from the beginning of the360 device.361 362 bitmap/chunksize363 The size, in bytes, of the chunk which will be represented by a364 single bit. For RAID456, it is a portion of an individual365 device. For RAID10, it is a portion of the array. For RAID1, it366 is both (they come to the same thing).367 368 bitmap/time_base369 The time, in seconds, between looking for bits in the bitmap to370 be cleared. In the current implementation, a bit will be cleared371 between 2 and 3 times ``time_base`` after all the covered blocks372 are known to be in-sync.373 374 bitmap/backlog375 When write-mostly devices are active in a RAID1, write requests376 to those devices proceed in the background - the filesystem (or377 other user of the device) does not have to wait for them.378 ``backlog`` sets a limit on the number of concurrent background379 writes. If there are more than this, new writes will by380 synchronous.381 382 bitmap/metadata383 This can be either ``internal`` or ``external``.384 385 ``internal``386 is the default and means the metadata for the bitmap387 is stored in the first 256 bytes of the allocated space and is388 managed by the md module.389 390 ``external``391 means that bitmap metadata is managed externally to392 the kernel (i.e. by some userspace program)393 394 bitmap/can_clear395 This is either ``true`` or ``false``. If ``true``, then bits in the396 bitmap will be cleared when the corresponding blocks are thought397 to be in-sync. If ``false``, bits will never be cleared.398 This is automatically set to ``false`` if a write happens on a399 degraded array, or if the array becomes degraded during a write.400 When metadata is managed externally, it should be set to true401 once the array becomes non-degraded, and this fact has been402 recorded in the metadata.403 404 consistency_policy405 This indicates how the array maintains consistency in case of unexpected406 shutdown. It can be:407 408 none409 Array has no redundancy information, e.g. raid0, linear.410 411 resync412 Full resync is performed and all redundancy is regenerated when the413 array is started after unclean shutdown.414 415 bitmap416 Resync assisted by a write-intent bitmap.417 418 journal419 For raid4/5/6, journal device is used to log transactions and replay420 after unclean shutdown.421 422 ppl423 For raid5 only, Partial Parity Log is used to close the write hole and424 eliminate resync.425 426 The accepted values when writing to this file are ``ppl`` and ``resync``,427 used to enable and disable PPL.428 429 uuid430 This indicates the UUID of the array in the following format:431 xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx432 433 434As component devices are added to an md array, they appear in the ``md``435directory as new directories named::436 437 dev-XXX438 439where ``XXX`` is a name that the kernel knows for the device, e.g. hdb1.440Each directory contains:441 442 block443 a symlink to the block device in /sys/block, e.g.::444 445 /sys/block/md0/md/dev-hdb1/block -> ../../../../block/hdb/hdb1446 447 super448 A file containing an image of the superblock read from, or449 written to, that device.450 451 state452 A file recording the current state of the device in the array453 which can be a comma separated list of:454 455 faulty456 device has been kicked from active use due to457 a detected fault, or it has unacknowledged bad458 blocks459 460 in_sync461 device is a fully in-sync member of the array462 463 writemostly464 device will only be subject to read465 requests if there are no other options.466 467 This applies only to raid1 arrays.468 469 blocked470 device has failed, and the failure hasn't been471 acknowledged yet by the metadata handler.472 473 Writes that would write to this device if474 it were not faulty are blocked.475 476 spare477 device is working, but not a full member.478 479 This includes spares that are in the process480 of being recovered to481 482 write_error483 device has ever seen a write error.484 485 want_replacement486 device is (mostly) working but probably487 should be replaced, either due to errors or488 due to user request.489 490 replacement491 device is a replacement for another active492 device with same raid_disk.493 494 495 This list may grow in future.496 497 This can be written to.498 499 Writing ``faulty`` simulates a failure on the device.500 501 Writing ``remove`` removes the device from the array.502 503 Writing ``writemostly`` sets the writemostly flag.504 505 Writing ``-writemostly`` clears the writemostly flag.506 507 Writing ``blocked`` sets the ``blocked`` flag.508 509 Writing ``-blocked`` clears the ``blocked`` flags and allows writes510 to complete and possibly simulates an error.511 512 Writing ``in_sync`` sets the in_sync flag.513 514 Writing ``write_error`` sets writeerrorseen flag.515 516 Writing ``-write_error`` clears writeerrorseen flag.517 518 Writing ``want_replacement`` is allowed at any time except to a519 replacement device or a spare. It sets the flag.520 521 Writing ``-want_replacement`` is allowed at any time. It clears522 the flag.523 524 Writing ``replacement`` or ``-replacement`` is only allowed before525 starting the array. It sets or clears the flag.526 527 528 This file responds to select/poll. Any change to ``faulty``529 or ``blocked`` causes an event.530 531 errors532 An approximate count of read errors that have been detected on533 this device but have not caused the device to be evicted from534 the array (either because they were corrected or because they535 happened while the array was read-only). When using version-1536 metadata, this value persists across restarts of the array.537 538 This value can be written while assembling an array thus539 providing an ongoing count for arrays with metadata managed by540 userspace.541 542 slot543 This gives the role that the device has in the array. It will544 either be ``none`` if the device is not active in the array545 (i.e. is a spare or has failed) or an integer less than the546 ``raid_disks`` number for the array indicating which position547 it currently fills. This can only be set while assembling an548 array. A device for which this is set is assumed to be working.549 550 offset551 This gives the location in the device (in sectors from the552 start) where data from the array will be stored. Any part of553 the device before this offset is not touched, unless it is554 used for storing metadata (Formats 1.1 and 1.2).555 556 size557 The amount of the device, after the offset, that can be used558 for storage of data. This will normally be the same as the559 component_size. This can be written while assembling an560 array. If a value less than the current component_size is561 written, it will be rejected.562 563 recovery_start564 When the device is not ``in_sync``, this records the number of565 sectors from the start of the device which are known to be566 correct. This is normally zero, but during a recovery567 operation it will steadily increase, and if the recovery is568 interrupted, restoring this value can cause recovery to569 avoid repeating the earlier blocks. With v1.x metadata, this570 value is saved and restored automatically.571 572 This can be set whenever the device is not an active member of573 the array, either before the array is activated, or before574 the ``slot`` is set.575 576 Setting this to ``none`` is equivalent to setting ``in_sync``.577 Setting to any other value also clears the ``in_sync`` flag.578 579 bad_blocks580 This gives the list of all known bad blocks in the form of581 start address and length (in sectors respectively). If output582 is too big to fit in a page, it will be truncated. Writing583 ``sector length`` to this file adds new acknowledged (i.e.584 recorded to disk safely) bad blocks.585 586 unacknowledged_bad_blocks587 This gives the list of known-but-not-yet-saved-to-disk bad588 blocks in the same form of ``bad_blocks``. If output is too big589 to fit in a page, it will be truncated. Writing to this file590 adds bad blocks without acknowledging them. This is largely591 for testing.592 593 ppl_sector, ppl_size594 Location and size (in sectors) of the space used for Partial Parity Log595 on this device.596 597 598An active md device will also contain an entry for each active device599in the array. These are named::600 601 rdNN602 603where ``NN`` is the position in the array, starting from 0.604So for a 3 drive array there will be rd0, rd1, rd2.605These are symbolic links to the appropriate ``dev-XXX`` entry.606Thus, for example::607 608 cat /sys/block/md*/md/rd*/state609 610will show ``in_sync`` on every line.611 612 613 614Active md devices for levels that support data redundancy (1,4,5,6,10)615also have616 617 sync_action618 a text file that can be used to monitor and control the rebuild619 process. It contains one word which can be one of:620 621 resync622 redundancy is being recalculated after unclean623 shutdown or creation624 625 recover626 a hot spare is being built to replace a627 failed/missing device628 629 idle630 nothing is happening631 check632 A full check of redundancy was requested and is633 happening. This reads all blocks and checks634 them. A repair may also happen for some raid635 levels.636 637 repair638 A full check and repair is happening. This is639 similar to ``resync``, but was requested by the640 user, and the write-intent bitmap is NOT used to641 optimise the process.642 643 This file is writable, and each of the strings that could be644 read are meaningful for writing.645 646 ``idle`` will stop an active resync/recovery etc. There is no647 guarantee that another resync/recovery may not be automatically648 started again, though some event will be needed to trigger649 this.650 651 ``resync`` or ``recovery`` can be used to restart the652 corresponding operation if it was stopped with ``idle``.653 654 ``check`` and ``repair`` will start the appropriate process655 providing the current state is ``idle``.656 657 This file responds to select/poll. Any important change in the value658 triggers a poll event. Sometimes the value will briefly be659 ``recover`` if a recovery seems to be needed, but cannot be660 achieved. In that case, the transition to ``recover`` isn't661 notified, but the transition away is.662 663 degraded664 This contains a count of the number of devices by which the665 arrays is degraded. So an optimal array will show ``0``. A666 single failed/missing drive will show ``1``, etc.667 668 This file responds to select/poll, any increase or decrease669 in the count of missing devices will trigger an event.670 671 mismatch_count672 When performing ``check`` and ``repair``, and possibly when673 performing ``resync``, md will count the number of errors that are674 found. The count in ``mismatch_cnt`` is the number of sectors675 that were re-written, or (for ``check``) would have been676 re-written. As most raid levels work in units of pages rather677 than sectors, this may be larger than the number of actual errors678 by a factor of the number of sectors in a page.679 680 bitmap_set_bits681 If the array has a write-intent bitmap, then writing to this682 attribute can set bits in the bitmap, indicating that a resync683 would need to check the corresponding blocks. Either individual684 numbers or start-end pairs can be written. Multiple numbers685 can be separated by a space.686 687 Note that the numbers are ``bit`` numbers, not ``block`` numbers.688 They should be scaled by the bitmap_chunksize.689 690 sync_speed_min, sync_speed_max691 This are similar to ``/proc/sys/dev/raid/speed_limit_{min,max}``692 however they only apply to the particular array.693 694 If no value has been written to these, or if the word ``system``695 is written, then the system-wide value is used. If a value,696 in kibibytes-per-second is written, then it is used.697 698 When the files are read, they show the currently active value699 followed by ``(local)`` or ``(system)`` depending on whether it is700 a locally set or system-wide value.701 702 sync_completed703 This shows the number of sectors that have been completed of704 whatever the current sync_action is, followed by the number of705 sectors in total that could need to be processed. The two706 numbers are separated by a ``/`` thus effectively showing one707 value, a fraction of the process that is complete.708 709 A ``select`` on this attribute will return when resync completes,710 when it reaches the current sync_max (below) and possibly at711 other times.712 713 sync_speed714 This shows the current actual speed, in K/sec, of the current715 sync_action. It is averaged over the last 30 seconds.716 717 suspend_lo, suspend_hi718 The two values, given as numbers of sectors, indicate a range719 within the array where IO will be blocked. This is currently720 only supported for raid4/5/6.721 722 sync_min, sync_max723 The two values, given as numbers of sectors, indicate a range724 within the array where ``check``/``repair`` will operate. Must be725 a multiple of chunk_size. When it reaches ``sync_max`` it will726 pause, rather than complete.727 You can use ``select`` or ``poll`` on ``sync_completed`` to wait for728 that number to reach sync_max. Then you can either increase729 ``sync_max``, or can write ``idle`` to ``sync_action``.730 731 The value of ``max`` for ``sync_max`` effectively disables the limit.732 When a resync is active, the value can only ever be increased,733 never decreased.734 The value of ``0`` is the minimum for ``sync_min``.735 736 737 738Each active md device may also have attributes specific to the739personality module that manages it.740These are specific to the implementation of the module and could741change substantially if the implementation changes.742 743These currently include:744 745 stripe_cache_size (currently raid5 only)746 number of entries in the stripe cache. This is writable, but747 there are upper and lower limits (32768, 17). Default is 256.748 749 strip_cache_active (currently raid5 only)750 number of active entries in the stripe cache751 752 preread_bypass_threshold (currently raid5 only)753 number of times a stripe requiring preread will be bypassed by754 a stripe that does not require preread. For fairness defaults755 to 1. Setting this to 0 disables bypass accounting and756 requires preread stripes to wait until all full-width stripe-757 writes are complete. Valid values are 0 to stripe_cache_size.758 759 journal_mode (currently raid5 only)760 The cache mode for raid5. raid5 could include an extra disk for761 caching. The mode can be "write-throuth" and "write-back". The762 default is "write-through".763 764 ppl_write_hint765 NVMe stream ID to be set for each PPL write request.766