Root/Documentation/md.txt

1Tools that manage md devices can be found at
2   http://www.kernel.org/pub/linux/utils/raid/
3
4
5Boot time assembly of RAID arrays
6---------------------------------
7
8You can boot with your md device with the following kernel command
9lines:
10
11for old raid arrays without persistent superblocks:
12  md=<md device no.>,<raid level>,<chunk size factor>,<fault level>,dev0,dev1,...,devn
13
14for raid arrays with persistent superblocks
15  md=<md device no.>,dev0,dev1,...,devn
16or, to assemble a partitionable array:
17  md=d<md device no.>,dev0,dev1,...,devn
18  
19md device no. = the number of the md device ...
20              0 means md0,
21          1 md1,
22          2 md2,
23          3 md3,
24          4 md4
25
26raid level = -1 linear mode
27              0 striped mode
28          other modes are only supported with persistent super blocks
29
30chunk size factor = (raid-0 and raid-1 only)
31              Set the chunk size as 4k << n.
32          
33fault level = totally ignored
34                
35dev0-devn: e.g. /dev/hda1,/dev/hdc1,/dev/sda1,/dev/sdb1
36                
37A possible loadlin line (Harald Hoyer <HarryH@Royal.Net>) looks like this:
38
39e:\loadlin\loadlin e:\zimage root=/dev/md0 md=0,0,4,0,/dev/hdb2,/dev/hdc3 ro
40
41
42Boot time autodetection of RAID arrays
43--------------------------------------
44
45When md is compiled into the kernel (not as module), partitions of
46type 0xfd are scanned and automatically assembled into RAID arrays.
47This autodetection may be suppressed with the kernel parameter
48"raid=noautodetect". As of kernel 2.6.9, only drives with a type 0
49superblock can be autodetected and run at boot time.
50
51The kernel parameter "raid=partitionable" (or "raid=part") means
52that all auto-detected arrays are assembled as partitionable.
53
54Boot time assembly of degraded/dirty arrays
55-------------------------------------------
56
57If a raid5 or raid6 array is both dirty and degraded, it could have
58undetectable data corruption. This is because the fact that it is
59'dirty' means that the parity cannot be trusted, and the fact that it
60is degraded means that some datablocks are missing and cannot reliably
61be reconstructed (due to no parity).
62
63For this reason, md will normally refuse to start such an array. This
64requires the sysadmin to take action to explicitly start the array
65despite possible corruption. This is normally done with
66   mdadm --assemble --force ....
67
68This option is not really available if the array has the root
69filesystem on it. In order to support this booting from such an
70array, md supports a module parameter "start_dirty_degraded" which,
71when set to 1, bypassed the checks and will allows dirty degraded
72arrays to be started.
73
74So, to boot with a root filesystem of a dirty degraded raid[56], use
75
76   md-mod.start_dirty_degraded=1
77
78
79Superblock formats
80------------------
81
82The md driver can support a variety of different superblock formats.
83Currently, it supports superblock formats "0.90.0" and the "md-1" format
84introduced in the 2.5 development series.
85
86The kernel will autodetect which format superblock is being used.
87
88Superblock format '0' is treated differently to others for legacy
89reasons - it is the original superblock format.
90
91
92General Rules - apply for all superblock formats
93------------------------------------------------
94
95An array is 'created' by writing appropriate superblocks to all
96devices.
97
98It is 'assembled' by associating each of these devices with an
99particular md virtual device. Once it is completely assembled, it can
100be accessed.
101
102An array should be created by a user-space tool. This will write
103superblocks to all devices. It will usually mark the array as
104'unclean', or with some devices missing so that the kernel md driver
105can create appropriate redundancy (copying in raid1, parity
106calculation in raid4/5).
107
108When an array is assembled, it is first initialized with the
109SET_ARRAY_INFO ioctl. This contains, in particular, a major and minor
110version number. The major version number selects which superblock
111format is to be used. The minor number might be used to tune handling
112of the format, such as suggesting where on each device to look for the
113superblock.
114
115Then each device is added using the ADD_NEW_DISK ioctl. This
116provides, in particular, a major and minor number identifying the
117device to add.
118
119The array is started with the RUN_ARRAY ioctl.
120
121Once started, new devices can be added. They should have an
122appropriate superblock written to them, and then passed be in with
123ADD_NEW_DISK.
124
125Devices that have failed or are not yet active can be detached from an
126array using HOT_REMOVE_DISK.
127
128
129Specific Rules that apply to format-0 super block arrays, and
130       arrays with no superblock (non-persistent).
131-------------------------------------------------------------
132
133An array can be 'created' by describing the array (level, chunksize
134etc) in a SET_ARRAY_INFO ioctl. This must has major_version==0 and
135raid_disks != 0.
136
137Then uninitialized devices can be added with ADD_NEW_DISK. The
138structure passed to ADD_NEW_DISK must specify the state of the device
139and its role in the array.
140
141Once started with RUN_ARRAY, uninitialized spares can be added with
142HOT_ADD_DISK.
143
144
145
146MD devices in sysfs
147-------------------
148md devices appear in sysfs (/sys) as regular block devices,
149e.g.
150   /sys/block/md0
151
152Each 'md' device will contain a subdirectory called 'md' which
153contains further md-specific information about the device.
154
155All md devices contain:
156  level
157     a text file indicating the 'raid level'. e.g. raid0, raid1,
158     raid5, linear, multipath, faulty.
159     If no raid level has been set yet (array is still being
160     assembled), the value will reflect whatever has been written
161     to it, which may be a name like the above, or may be a number
162     such as '0', '5', etc.
163
164  raid_disks
165     a text file with a simple number indicating the number of devices
166     in a fully functional array. If this is not yet known, the file
167     will be empty. If an array is being resized this will contain
168     the new number of devices.
169     Some raid levels allow this value to be set while the array is
170     active. This will reconfigure the array. Otherwise it can only
171     be set while assembling an array.
172     A change to this attribute will not be permitted if it would
173     reduce the size of the array. To reduce the number of drives
174     in an e.g. raid5, the array size must first be reduced by
175     setting the 'array_size' attribute.
176
177  chunk_size
178     This is the size in bytes for 'chunks' and is only relevant to
179     raid levels that involve striping (0,4,5,6,10). The address space
180     of the array is conceptually divided into chunks and consecutive
181     chunks are striped onto neighbouring devices.
182     The size should be at least PAGE_SIZE (4k) and should be a power
183     of 2. This can only be set while assembling an array
184
185  layout
186     The "layout" for the array for the particular level. This is
187     simply a number that is interpretted differently by different
188     levels. It can be written while assembling an array.
189
190  array_size
191     This can be used to artificially constrain the available space in
192     the array to be less than is actually available on the combined
193     devices. Writing a number (in Kilobytes) which is less than
194     the available size will set the size. Any reconfiguration of the
195     array (e.g. adding devices) will not cause the size to change.
196     Writing the word 'default' will cause the effective size of the
197     array to be whatever size is actually available based on
198     'level', 'chunk_size' and 'component_size'.
199
200     This can be used to reduce the size of the array before reducing
201     the number of devices in a raid4/5/6, or to support external
202     metadata formats which mandate such clipping.
203
204  reshape_position
205     This is either "none" or a sector number within the devices of
206     the array where "reshape" is up to. If this is set, the three
207     attributes mentioned above (raid_disks, chunk_size, layout) can
208     potentially have 2 values, an old and a new value. If these
209     values differ, reading the attribute returns
210        new (old)
211     and writing will effect the 'new' value, leaving the 'old'
212     unchanged.
213
214  component_size
215     For arrays with data redundancy (i.e. not raid0, linear, faulty,
216     multipath), all components must be the same size - or at least
217     there must a size that they all provide space for. This is a key
218     part or the geometry of the array. It is measured in sectors
219     and can be read from here. Writing to this value may resize
220     the array if the personality supports it (raid1, raid5, raid6),
221     and if the component drives are large enough.
222
223  metadata_version
224     This indicates the format that is being used to record metadata
225     about the array. It can be 0.90 (traditional format), 1.0, 1.1,
226     1.2 (newer format in varying locations) or "none" indicating that
227     the kernel isn't managing metadata at all.
228     Alternately it can be "external:" followed by a string which
229     is set by user-space. This indicates that metadata is managed
230     by a user-space program. Any device failure or other event that
231     requires a metadata update will cause array activity to be
232     suspended until the event is acknowledged.
233
234  resync_start
235     The point at which resync should start. If no resync is needed,
236     this will be a very large number (or 'none' since 2.6.30-rc1). At
237     array creation it will default to 0, though starting the array as
238     'clean' will set it much larger.
239
240   new_dev
241     This file can be written but not read. The value written should
242     be a block device number as major:minor. e.g. 8:0
243     This will cause that device to be attached to the array, if it is
244     available. It will then appear at md/dev-XXX (depending on the
245     name of the device) and further configuration is then possible.
246
247   safe_mode_delay
248     When an md array has seen no write requests for a certain period
249     of time, it will be marked as 'clean'. When another write
250     request arrives, the array is marked as 'dirty' before the write
251     commences. This is known as 'safe_mode'.
252     The 'certain period' is controlled by this file which stores the
253     period as a number of seconds. The default is 200msec (0.200).
254     Writing a value of 0 disables safemode.
255
256   array_state
257     This file contains a single word which describes the current
258     state of the array. In many cases, the state can be set by
259     writing the word for the desired state, however some states
260     cannot be explicitly set, and some transitions are not allowed.
261
262     Select/poll works on this file. All changes except between
263         active_idle and active (which can be frequent and are not
264    very interesting) are notified. active->active_idle is
265    reported if the metadata is externally managed.
266
267     clear
268         No devices, no size, no level
269         Writing is equivalent to STOP_ARRAY ioctl
270     inactive
271         May have some settings, but array is not active
272            all IO results in error
273         When written, doesn't tear down array, but just stops it
274     suspended (not supported yet)
275         All IO requests will block. The array can be reconfigured.
276         Writing this, if accepted, will block until array is quiessent
277     readonly
278         no resync can happen. no superblocks get written.
279         write requests fail
280     read-auto
281         like readonly, but behaves like 'clean' on a write request.
282
283     clean - no pending writes, but otherwise active.
284         When written to inactive array, starts without resync
285         If a write request arrives then
286           if metadata is known, mark 'dirty' and switch to 'active'.
287           if not known, block and switch to write-pending
288         If written to an active array that has pending writes, then fails.
289     active
290         fully active: IO and resync can be happening.
291         When written to inactive array, starts with resync
292
293     write-pending
294         clean, but writes are blocked waiting for 'active' to be written.
295
296     active-idle
297         like active, but no writes have been seen for a while (safe_mode_delay).
298
299  bitmap/location
300     This indicates where the write-intent bitmap for the array is
301     stored.
302     It can be one of "none", "file" or "[+-]N".
303     "file" may later be extended to "file:/file/name"
304     "[+-]N" means that many sectors from the start of the metadata.
305       This is replicated on all devices. For arrays with externally
306       managed metadata, the offset is from the beginning of the
307       device.
308  bitmap/chunksize
309     The size, in bytes, of the chunk which will be represented by a
310     single bit. For RAID456, it is a portion of an individual
311     device. For RAID10, it is a portion of the array. For RAID1, it
312     is both (they come to the same thing).
313  bitmap/time_base
314     The time, in seconds, between looking for bits in the bitmap to
315     be cleared. In the current implementation, a bit will be cleared
316     between 2 and 3 times "time_base" after all the covered blocks
317     are known to be in-sync.
318  bitmap/backlog
319     When write-mostly devices are active in a RAID1, write requests
320     to those devices proceed in the background - the filesystem (or
321     other user of the device) does not have to wait for them.
322     'backlog' sets a limit on the number of concurrent background
323     writes. If there are more than this, new writes will by
324     synchronous.
325  bitmap/metadata
326     This can be either 'internal' or 'external'.
327     'internal' is the default and means the metadata for the bitmap
328     is stored in the first 256 bytes of the allocated space and is
329     managed by the md module.
330     'external' means that bitmap metadata is managed externally to
331     the kernel (i.e. by some userspace program)
332  bitmap/can_clear
333     This is either 'true' or 'false'. If 'true', then bits in the
334     bitmap will be cleared when the corresponding blocks are thought
335     to be in-sync. If 'false', bits will never be cleared.
336     This is automatically set to 'false' if a write happens on a
337     degraded array, or if the array becomes degraded during a write.
338     When metadata is managed externally, it should be set to true
339     once the array becomes non-degraded, and this fact has been
340     recorded in the metadata.
341     
342     
343     
344
345As component devices are added to an md array, they appear in the 'md'
346directory as new directories named
347      dev-XXX
348where XXX is a name that the kernel knows for the device, e.g. hdb1.
349Each directory contains:
350
351      block
352        a symlink to the block device in /sys/block, e.g.
353         /sys/block/md0/md/dev-hdb1/block -> ../../../../block/hdb/hdb1
354
355      super
356        A file containing an image of the superblock read from, or
357        written to, that device.
358
359      state
360        A file recording the current state of the device in the array
361    which can be a comma separated list of
362          faulty - device has been kicked from active use due to
363                         a detected fault or it has unacknowledged bad
364                         blocks
365          in_sync - device is a fully in-sync member of the array
366          writemostly - device will only be subject to read
367                 requests if there are no other options.
368             This applies only to raid1 arrays.
369          blocked - device has failed, and the failure hasn't been
370             acknowledged yet by the metadata handler.
371             Writes that would write to this device if
372             it were not faulty are blocked.
373          spare - device is working, but not a full member.
374             This includes spares that are in the process
375             of being recovered to
376          write_error - device has ever seen a write error.
377    This list may grow in future.
378    This can be written to.
379    Writing "faulty" simulates a failure on the device.
380    Writing "remove" removes the device from the array.
381    Writing "writemostly" sets the writemostly flag.
382    Writing "-writemostly" clears the writemostly flag.
383    Writing "blocked" sets the "blocked" flag.
384    Writing "-blocked" clears the "blocked" flags and allows writes
385        to complete and possibly simulates an error.
386    Writing "in_sync" sets the in_sync flag.
387    Writing "write_error" sets writeerrorseen flag.
388    Writing "-write_error" clears writeerrorseen flag.
389
390    This file responds to select/poll. Any change to 'faulty'
391    or 'blocked' causes an event.
392
393      errors
394    An approximate count of read errors that have been detected on
395    this device but have not caused the device to be evicted from
396    the array (either because they were corrected or because they
397    happened while the array was read-only). When using version-1
398    metadata, this value persists across restarts of the array.
399
400    This value can be written while assembling an array thus
401    providing an ongoing count for arrays with metadata managed by
402    userspace.
403
404      slot
405        This gives the role that the device has in the array. It will
406    either be 'none' if the device is not active in the array
407        (i.e. is a spare or has failed) or an integer less than the
408    'raid_disks' number for the array indicating which position
409    it currently fills. This can only be set while assembling an
410    array. A device for which this is set is assumed to be working.
411
412      offset
413        This gives the location in the device (in sectors from the
414        start) where data from the array will be stored. Any part of
415        the device before this offset us not touched, unless it is
416        used for storing metadata (Formats 1.1 and 1.2).
417
418      size
419        The amount of the device, after the offset, that can be used
420        for storage of data. This will normally be the same as the
421    component_size. This can be written while assembling an
422        array. If a value less than the current component_size is
423        written, it will be rejected.
424
425      recovery_start
426        When the device is not 'in_sync', this records the number of
427    sectors from the start of the device which are known to be
428    correct. This is normally zero, but during a recovery
429    operation is will steadily increase, and if the recovery is
430    interrupted, restoring this value can cause recovery to
431    avoid repeating the earlier blocks. With v1.x metadata, this
432    value is saved and restored automatically.
433
434    This can be set whenever the device is not an active member of
435    the array, either before the array is activated, or before
436    the 'slot' is set.
437
438    Setting this to 'none' is equivalent to setting 'in_sync'.
439    Setting to any other value also clears the 'in_sync' flag.
440    
441      bad_blocks
442    This gives the list of all known bad blocks in the form of
443    start address and length (in sectors respectively). If output
444    is too big to fit in a page, it will be truncated. Writing
445    "sector length" to this file adds new acknowledged (i.e.
446    recorded to disk safely) bad blocks.
447
448      unacknowledged_bad_blocks
449    This gives the list of known-but-not-yet-saved-to-disk bad
450    blocks in the same form of 'bad_blocks'. If output is too big
451    to fit in a page, it will be truncated. Writing to this file
452    adds bad blocks without acknowledging them. This is largely
453    for testing.
454
455
456
457An active md device will also contain and entry for each active device
458in the array. These are named
459
460    rdNN
461
462where 'NN' is the position in the array, starting from 0.
463So for a 3 drive array there will be rd0, rd1, rd2.
464These are symbolic links to the appropriate 'dev-XXX' entry.
465Thus, for example,
466       cat /sys/block/md*/md/rd*/state
467will show 'in_sync' on every line.
468
469
470
471Active md devices for levels that support data redundancy (1,4,5,6)
472also have
473
474   sync_action
475     a text file that can be used to monitor and control the rebuild
476     process. It contains one word which can be one of:
477       resync - redundancy is being recalculated after unclean
478                       shutdown or creation
479       recover - a hot spare is being built to replace a
480                       failed/missing device
481       idle - nothing is happening
482       check - A full check of redundancy was requested and is
483                       happening. This reads all block and checks
484                       them. A repair may also happen for some raid
485                       levels.
486       repair - A full check and repair is happening. This is
487                       similar to 'resync', but was requested by the
488                       user, and the write-intent bitmap is NOT used to
489               optimise the process.
490
491      This file is writable, and each of the strings that could be
492      read are meaningful for writing.
493
494       'idle' will stop an active resync/recovery etc. There is no
495           guarantee that another resync/recovery may not be automatically
496       started again, though some event will be needed to trigger
497           this.
498    'resync' or 'recovery' can be used to restart the
499           corresponding operation if it was stopped with 'idle'.
500    'check' and 'repair' will start the appropriate process
501           providing the current state is 'idle'.
502
503      This file responds to select/poll. Any important change in the value
504      triggers a poll event. Sometimes the value will briefly be
505      "recover" if a recovery seems to be needed, but cannot be
506      achieved. In that case, the transition to "recover" isn't
507      notified, but the transition away is.
508
509   degraded
510      This contains a count of the number of devices by which the
511      arrays is degraded. So an optimal array with show '0'. A
512      single failed/missing drive will show '1', etc.
513      This file responds to select/poll, any increase or decrease
514      in the count of missing devices will trigger an event.
515
516   mismatch_count
517      When performing 'check' and 'repair', and possibly when
518      performing 'resync', md will count the number of errors that are
519      found. The count in 'mismatch_cnt' is the number of sectors
520      that were re-written, or (for 'check') would have been
521      re-written. As most raid levels work in units of pages rather
522      than sectors, this my be larger than the number of actual errors
523      by a factor of the number of sectors in a page.
524
525   bitmap_set_bits
526      If the array has a write-intent bitmap, then writing to this
527      attribute can set bits in the bitmap, indicating that a resync
528      would need to check the corresponding blocks. Either individual
529      numbers or start-end pairs can be written. Multiple numbers
530      can be separated by a space.
531      Note that the numbers are 'bit' numbers, not 'block' numbers.
532      They should be scaled by the bitmap_chunksize.
533
534   sync_speed_min
535   sync_speed_max
536     This are similar to /proc/sys/dev/raid/speed_limit_{min,max}
537     however they only apply to the particular array.
538     If no value has been written to these, of if the word 'system'
539     is written, then the system-wide value is used. If a value,
540     in kibibytes-per-second is written, then it is used.
541     When the files are read, they show the currently active value
542     followed by "(local)" or "(system)" depending on whether it is
543     a locally set or system-wide value.
544
545   sync_completed
546     This shows the number of sectors that have been completed of
547     whatever the current sync_action is, followed by the number of
548     sectors in total that could need to be processed. The two
549     numbers are separated by a '/' thus effectively showing one
550     value, a fraction of the process that is complete.
551     A 'select' on this attribute will return when resync completes,
552     when it reaches the current sync_max (below) and possibly at
553     other times.
554
555   sync_max
556     This is a number of sectors at which point a resync/recovery
557     process will pause. When a resync is active, the value can
558     only ever be increased, never decreased. The value of 'max'
559     effectively disables the limit.
560
561
562   sync_speed
563     This shows the current actual speed, in K/sec, of the current
564     sync_action. It is averaged over the last 30 seconds.
565
566   suspend_lo
567   suspend_hi
568     The two values, given as numbers of sectors, indicate a range
569     within the array where IO will be blocked. This is currently
570     only supported for raid4/5/6.
571
572   sync_min
573   sync_max
574     The two values, given as numbers of sectors, indicate a range
575     within the array where 'check'/'repair' will operate. Must be
576     a multiple of chunk_size. When it reaches "sync_max" it will
577     pause, rather than complete.
578     You can use 'select' or 'poll' on "sync_completed" to wait for
579     that number to reach sync_max. Then you can either increase
580     "sync_max", or can write 'idle' to "sync_action".
581
582
583Each active md device may also have attributes specific to the
584personality module that manages it.
585These are specific to the implementation of the module and could
586change substantially if the implementation changes.
587
588These currently include
589
590  stripe_cache_size (currently raid5 only)
591      number of entries in the stripe cache. This is writable, but
592      there are upper and lower limits (32768, 16). Default is 128.
593  strip_cache_active (currently raid5 only)
594      number of active entries in the stripe cache
595  preread_bypass_threshold (currently raid5 only)
596      number of times a stripe requiring preread will be bypassed by
597      a stripe that does not require preread. For fairness defaults
598      to 1. Setting this to 0 disables bypass accounting and
599      requires preread stripes to wait until all full-width stripe-
600      writes are complete. Valid values are 0 to stripe_cache_size.
601

Archive Download this file



interactive