md.txt 17 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426
  1. Tools that manage md devices can be found at
  2. http://www.<country>.kernel.org/pub/linux/utils/raid/....
  3. Boot time assembly of RAID arrays
  4. ---------------------------------
  5. You can boot with your md device with the following kernel command
  6. lines:
  7. for old raid arrays without persistent superblocks:
  8. md=<md device no.>,<raid level>,<chunk size factor>,<fault level>,dev0,dev1,...,devn
  9. for raid arrays with persistent superblocks
  10. md=<md device no.>,dev0,dev1,...,devn
  11. or, to assemble a partitionable array:
  12. md=d<md device no.>,dev0,dev1,...,devn
  13. md device no. = the number of the md device ...
  14. 0 means md0,
  15. 1 md1,
  16. 2 md2,
  17. 3 md3,
  18. 4 md4
  19. raid level = -1 linear mode
  20. 0 striped mode
  21. other modes are only supported with persistent super blocks
  22. chunk size factor = (raid-0 and raid-1 only)
  23. Set the chunk size as 4k << n.
  24. fault level = totally ignored
  25. dev0-devn: e.g. /dev/hda1,/dev/hdc1,/dev/sda1,/dev/sdb1
  26. A possible loadlin line (Harald Hoyer <HarryH@Royal.Net>) looks like this:
  27. e:\loadlin\loadlin e:\zimage root=/dev/md0 md=0,0,4,0,/dev/hdb2,/dev/hdc3 ro
  28. Boot time autodetection of RAID arrays
  29. --------------------------------------
  30. When md is compiled into the kernel (not as module), partitions of
  31. type 0xfd are scanned and automatically assembled into RAID arrays.
  32. This autodetection may be suppressed with the kernel parameter
  33. "raid=noautodetect". As of kernel 2.6.9, only drives with a type 0
  34. superblock can be autodetected and run at boot time.
  35. The kernel parameter "raid=partitionable" (or "raid=part") means
  36. that all auto-detected arrays are assembled as partitionable.
  37. Boot time assembly of degraded/dirty arrays
  38. -------------------------------------------
  39. If a raid5 or raid6 array is both dirty and degraded, it could have
  40. undetectable data corruption. This is because the fact that it is
  41. 'dirty' means that the parity cannot be trusted, and the fact that it
  42. is degraded means that some datablocks are missing and cannot reliably
  43. be reconstructed (due to no parity).
  44. For this reason, md will normally refuse to start such an array. This
  45. requires the sysadmin to take action to explicitly start the array
  46. despite possible corruption. This is normally done with
  47. mdadm --assemble --force ....
  48. This option is not really available if the array has the root
  49. filesystem on it. In order to support this booting from such an
  50. array, md supports a module parameter "start_dirty_degraded" which,
  51. when set to 1, bypassed the checks and will allows dirty degraded
  52. arrays to be started.
  53. So, to boot with a root filesystem of a dirty degraded raid[56], use
  54. md-mod.start_dirty_degraded=1
  55. Superblock formats
  56. ------------------
  57. The md driver can support a variety of different superblock formats.
  58. Currently, it supports superblock formats "0.90.0" and the "md-1" format
  59. introduced in the 2.5 development series.
  60. The kernel will autodetect which format superblock is being used.
  61. Superblock format '0' is treated differently to others for legacy
  62. reasons - it is the original superblock format.
  63. General Rules - apply for all superblock formats
  64. ------------------------------------------------
  65. An array is 'created' by writing appropriate superblocks to all
  66. devices.
  67. It is 'assembled' by associating each of these devices with an
  68. particular md virtual device. Once it is completely assembled, it can
  69. be accessed.
  70. An array should be created by a user-space tool. This will write
  71. superblocks to all devices. It will usually mark the array as
  72. 'unclean', or with some devices missing so that the kernel md driver
  73. can create appropriate redundancy (copying in raid1, parity
  74. calculation in raid4/5).
  75. When an array is assembled, it is first initialized with the
  76. SET_ARRAY_INFO ioctl. This contains, in particular, a major and minor
  77. version number. The major version number selects which superblock
  78. format is to be used. The minor number might be used to tune handling
  79. of the format, such as suggesting where on each device to look for the
  80. superblock.
  81. Then each device is added using the ADD_NEW_DISK ioctl. This
  82. provides, in particular, a major and minor number identifying the
  83. device to add.
  84. The array is started with the RUN_ARRAY ioctl.
  85. Once started, new devices can be added. They should have an
  86. appropriate superblock written to them, and then passed be in with
  87. ADD_NEW_DISK.
  88. Devices that have failed or are not yet active can be detached from an
  89. array using HOT_REMOVE_DISK.
  90. Specific Rules that apply to format-0 super block arrays, and
  91. arrays with no superblock (non-persistent).
  92. -------------------------------------------------------------
  93. An array can be 'created' by describing the array (level, chunksize
  94. etc) in a SET_ARRAY_INFO ioctl. This must has major_version==0 and
  95. raid_disks != 0.
  96. Then uninitialized devices can be added with ADD_NEW_DISK. The
  97. structure passed to ADD_NEW_DISK must specify the state of the device
  98. and it's role in the array.
  99. Once started with RUN_ARRAY, uninitialized spares can be added with
  100. HOT_ADD_DISK.
  101. MD devices in sysfs
  102. -------------------
  103. md devices appear in sysfs (/sys) as regular block devices,
  104. e.g.
  105. /sys/block/md0
  106. Each 'md' device will contain a subdirectory called 'md' which
  107. contains further md-specific information about the device.
  108. All md devices contain:
  109. level
  110. a text file indicating the 'raid level'. e.g. raid0, raid1,
  111. raid5, linear, multipath, faulty.
  112. If no raid level has been set yet (array is still being
  113. assembled), the value will reflect whatever has been written
  114. to it, which may be a name like the above, or may be a number
  115. such as '0', '5', etc.
  116. raid_disks
  117. a text file with a simple number indicating the number of devices
  118. in a fully functional array. If this is not yet known, the file
  119. will be empty. If an array is being resized (not currently
  120. possible) this will contain the larger of the old and new sizes.
  121. Some raid level (RAID1) allow this value to be set while the
  122. array is active. This will reconfigure the array. Otherwise
  123. it can only be set while assembling an array.
  124. chunk_size
  125. This is the size if bytes for 'chunks' and is only relevant to
  126. raid levels that involve striping (1,4,5,6,10). The address space
  127. of the array is conceptually divided into chunks and consecutive
  128. chunks are striped onto neighbouring devices.
  129. The size should be at least PAGE_SIZE (4k) and should be a power
  130. of 2. This can only be set while assembling an array
  131. component_size
  132. For arrays with data redundancy (i.e. not raid0, linear, faulty,
  133. multipath), all components must be the same size - or at least
  134. there must a size that they all provide space for. This is a key
  135. part or the geometry of the array. It is measured in sectors
  136. and can be read from here. Writing to this value may resize
  137. the array if the personality supports it (raid1, raid5, raid6),
  138. and if the component drives are large enough.
  139. metadata_version
  140. This indicates the format that is being used to record metadata
  141. about the array. It can be 0.90 (traditional format), 1.0, 1.1,
  142. 1.2 (newer format in varying locations) or "none" indicating that
  143. the kernel isn't managing metadata at all.
  144. layout
  145. The "layout" for the array for the particular level. This is
  146. simply a number that is interpretted differently by different
  147. levels. It can be written while assembling an array.
  148. resync_start
  149. The point at which resync should start. If no resync is needed,
  150. this will be a very large number. At array creation it will
  151. default to 0, though starting the array as 'clean' will
  152. set it much larger.
  153. new_dev
  154. This file can be written but not read. The value written should
  155. be a block device number as major:minor. e.g. 8:0
  156. This will cause that device to be attached to the array, if it is
  157. available. It will then appear at md/dev-XXX (depending on the
  158. name of the device) and further configuration is then possible.
  159. safe_mode_delay
  160. When an md array has seen no write requests for a certain period
  161. of time, it will be marked as 'clean'. When another write
  162. request arrives, the array is marked as 'dirty' before the write
  163. commences. This is known as 'safe_mode'.
  164. The 'certain period' is controlled by this file which stores the
  165. period as a number of seconds. The default is 200msec (0.200).
  166. Writing a value of 0 disables safemode.
  167. array_state
  168. This file contains a single word which describes the current
  169. state of the array. In many cases, the state can be set by
  170. writing the word for the desired state, however some states
  171. cannot be explicitly set, and some transitions are not allowed.
  172. clear
  173. No devices, no size, no level
  174. Writing is equivalent to STOP_ARRAY ioctl
  175. inactive
  176. May have some settings, but array is not active
  177. all IO results in error
  178. When written, doesn't tear down array, but just stops it
  179. suspended (not supported yet)
  180. All IO requests will block. The array can be reconfigured.
  181. Writing this, if accepted, will block until array is quiessent
  182. readonly
  183. no resync can happen. no superblocks get written.
  184. write requests fail
  185. read-auto
  186. like readonly, but behaves like 'clean' on a write request.
  187. clean - no pending writes, but otherwise active.
  188. When written to inactive array, starts without resync
  189. If a write request arrives then
  190. if metadata is known, mark 'dirty' and switch to 'active'.
  191. if not known, block and switch to write-pending
  192. If written to an active array that has pending writes, then fails.
  193. active
  194. fully active: IO and resync can be happening.
  195. When written to inactive array, starts with resync
  196. write-pending
  197. clean, but writes are blocked waiting for 'active' to be written.
  198. active-idle
  199. like active, but no writes have been seen for a while (safe_mode_delay).
  200. sync_speed_min
  201. sync_speed_max
  202. This are similar to /proc/sys/dev/raid/speed_limit_{min,max}
  203. however they only apply to the particular array.
  204. If no value has been written to these, of if the word 'system'
  205. is written, then the system-wide value is used. If a value,
  206. in kibibytes-per-second is written, then it is used.
  207. When the files are read, they show the currently active value
  208. followed by "(local)" or "(system)" depending on whether it is
  209. a locally set or system-wide value.
  210. sync_completed
  211. This shows the number of sectors that have been completed of
  212. whatever the current sync_action is, followed by the number of
  213. sectors in total that could need to be processed. The two
  214. numbers are separated by a '/' thus effectively showing one
  215. value, a fraction of the process that is complete.
  216. sync_speed
  217. This shows the current actual speed, in K/sec, of the current
  218. sync_action. It is averaged over the last 30 seconds.
  219. As component devices are added to an md array, they appear in the 'md'
  220. directory as new directories named
  221. dev-XXX
  222. where XXX is a name that the kernel knows for the device, e.g. hdb1.
  223. Each directory contains:
  224. block
  225. a symlink to the block device in /sys/block, e.g.
  226. /sys/block/md0/md/dev-hdb1/block -> ../../../../block/hdb/hdb1
  227. super
  228. A file containing an image of the superblock read from, or
  229. written to, that device.
  230. state
  231. A file recording the current state of the device in the array
  232. which can be a comma separated list of
  233. faulty - device has been kicked from active use due to
  234. a detected fault
  235. in_sync - device is a fully in-sync member of the array
  236. writemostly - device will only be subject to read
  237. requests if there are no other options.
  238. This applies only to raid1 arrays.
  239. spare - device is working, but not a full member.
  240. This includes spares that are in the process
  241. of being recovered to
  242. This list may grow in future.
  243. This can be written to.
  244. Writing "faulty" simulates a failure on the device.
  245. Writing "remove" removes the device from the array.
  246. Writing "writemostly" sets the writemostly flag.
  247. Writing "-writemostly" clears the writemostly flag.
  248. errors
  249. An approximate count of read errors that have been detected on
  250. this device but have not caused the device to be evicted from
  251. the array (either because they were corrected or because they
  252. happened while the array was read-only). When using version-1
  253. metadata, this value persists across restarts of the array.
  254. This value can be written while assembling an array thus
  255. providing an ongoing count for arrays with metadata managed by
  256. userspace.
  257. slot
  258. This gives the role that the device has in the array. It will
  259. either be 'none' if the device is not active in the array
  260. (i.e. is a spare or has failed) or an integer less than the
  261. 'raid_disks' number for the array indicating which position
  262. it currently fills. This can only be set while assembling an
  263. array. A device for which this is set is assumed to be working.
  264. offset
  265. This gives the location in the device (in sectors from the
  266. start) where data from the array will be stored. Any part of
  267. the device before this offset us not touched, unless it is
  268. used for storing metadata (Formats 1.1 and 1.2).
  269. size
  270. The amount of the device, after the offset, that can be used
  271. for storage of data. This will normally be the same as the
  272. component_size. This can be written while assembling an
  273. array. If a value less than the current component_size is
  274. written, component_size will be reduced to this value.
  275. An active md device will also contain and entry for each active device
  276. in the array. These are named
  277. rdNN
  278. where 'NN' is the position in the array, starting from 0.
  279. So for a 3 drive array there will be rd0, rd1, rd2.
  280. These are symbolic links to the appropriate 'dev-XXX' entry.
  281. Thus, for example,
  282. cat /sys/block/md*/md/rd*/state
  283. will show 'in_sync' on every line.
  284. Active md devices for levels that support data redundancy (1,4,5,6)
  285. also have
  286. sync_action
  287. a text file that can be used to monitor and control the rebuild
  288. process. It contains one word which can be one of:
  289. resync - redundancy is being recalculated after unclean
  290. shutdown or creation
  291. recover - a hot spare is being built to replace a
  292. failed/missing device
  293. idle - nothing is happening
  294. check - A full check of redundancy was requested and is
  295. happening. This reads all block and checks
  296. them. A repair may also happen for some raid
  297. levels.
  298. repair - A full check and repair is happening. This is
  299. similar to 'resync', but was requested by the
  300. user, and the write-intent bitmap is NOT used to
  301. optimise the process.
  302. This file is writable, and each of the strings that could be
  303. read are meaningful for writing.
  304. 'idle' will stop an active resync/recovery etc. There is no
  305. guarantee that another resync/recovery may not be automatically
  306. started again, though some event will be needed to trigger
  307. this.
  308. 'resync' or 'recovery' can be used to restart the
  309. corresponding operation if it was stopped with 'idle'.
  310. 'check' and 'repair' will start the appropriate process
  311. providing the current state is 'idle'.
  312. mismatch_count
  313. When performing 'check' and 'repair', and possibly when
  314. performing 'resync', md will count the number of errors that are
  315. found. The count in 'mismatch_cnt' is the number of sectors
  316. that were re-written, or (for 'check') would have been
  317. re-written. As most raid levels work in units of pages rather
  318. than sectors, this my be larger than the number of actual errors
  319. by a factor of the number of sectors in a page.
  320. bitmap_set_bits
  321. If the array has a write-intent bitmap, then writing to this
  322. attribute can set bits in the bitmap, indicating that a resync
  323. would need to check the corresponding blocks. Either individual
  324. numbers or start-end pairs can be written. Multiple numbers
  325. can be separated by a space.
  326. Note that the numbers are 'bit' numbers, not 'block' numbers.
  327. They should be scaled by the bitmap_chunksize.
  328. Each active md device may also have attributes specific to the
  329. personality module that manages it.
  330. These are specific to the implementation of the module and could
  331. change substantially if the implementation changes.
  332. These currently include
  333. stripe_cache_size (currently raid5 only)
  334. number of entries in the stripe cache. This is writable, but
  335. there are upper and lower limits (32768, 16). Default is 128.
  336. strip_cache_active (currently raid5 only)
  337. number of active entries in the stripe cache