annotate Documentation/md.txt @ 897:329ea0ccb344

balloon: try harder to balloon up under memory pressure.

Currently if the balloon driver is unable to increase the guest's
reservation it assumes the failure was due to reaching its full
allocation, gives up on the ballooning operation and records the limit
it reached as the "hard limit". The driver will not try again until
the target is set again (even to the same value).

However it is possible that ballooning has in fact failed due to
memory pressure in the host and therefore it is desirable to keep
attempting to reach the target in case memory becomes available. The
most likely scenario is that some guests are ballooning down while
others are ballooning up and therefore there is temporary memory
pressure while things stabilise. You would not expect a well behaved
toolstack to ask a domain to balloon to more than its allocation nor
would you expect it to deliberately over-commit memory by setting
balloon targets which exceed the total host memory.

This patch drops the concept of a hard limit and causes the balloon
driver to retry increasing the reservation on a timer in the same
manner as when decreasing the reservation.

Also if we partially succeed in increasing the reservation
(i.e. receive less pages than we asked for) then we may as well keep
those pages rather than returning them to Xen.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
author Keir Fraser <keir.fraser@citrix.com>
date Fri Jun 05 14:01:20 2009 +0100 (2009-06-05)
parents 831230e53067
rev   line source
ian@0 1 Tools that manage md devices can be found at
ian@0 2 http://www.<country>.kernel.org/pub/linux/utils/raid/....
ian@0 3
ian@0 4
ian@0 5 Boot time assembly of RAID arrays
ian@0 6 ---------------------------------
ian@0 7
ian@0 8 You can boot with your md device with the following kernel command
ian@0 9 lines:
ian@0 10
ian@0 11 for old raid arrays without persistent superblocks:
ian@0 12 md=<md device no.>,<raid level>,<chunk size factor>,<fault level>,dev0,dev1,...,devn
ian@0 13
ian@0 14 for raid arrays with persistent superblocks
ian@0 15 md=<md device no.>,dev0,dev1,...,devn
ian@0 16 or, to assemble a partitionable array:
ian@0 17 md=d<md device no.>,dev0,dev1,...,devn
ian@0 18
ian@0 19 md device no. = the number of the md device ...
ian@0 20 0 means md0,
ian@0 21 1 md1,
ian@0 22 2 md2,
ian@0 23 3 md3,
ian@0 24 4 md4
ian@0 25
ian@0 26 raid level = -1 linear mode
ian@0 27 0 striped mode
ian@0 28 other modes are only supported with persistent super blocks
ian@0 29
ian@0 30 chunk size factor = (raid-0 and raid-1 only)
ian@0 31 Set the chunk size as 4k << n.
ian@0 32
ian@0 33 fault level = totally ignored
ian@0 34
ian@0 35 dev0-devn: e.g. /dev/hda1,/dev/hdc1,/dev/sda1,/dev/sdb1
ian@0 36
ian@0 37 A possible loadlin line (Harald Hoyer <HarryH@Royal.Net>) looks like this:
ian@0 38
ian@0 39 e:\loadlin\loadlin e:\zimage root=/dev/md0 md=0,0,4,0,/dev/hdb2,/dev/hdc3 ro
ian@0 40
ian@0 41
ian@0 42 Boot time autodetection of RAID arrays
ian@0 43 --------------------------------------
ian@0 44
ian@0 45 When md is compiled into the kernel (not as module), partitions of
ian@0 46 type 0xfd are scanned and automatically assembled into RAID arrays.
ian@0 47 This autodetection may be suppressed with the kernel parameter
ian@0 48 "raid=noautodetect". As of kernel 2.6.9, only drives with a type 0
ian@0 49 superblock can be autodetected and run at boot time.
ian@0 50
ian@0 51 The kernel parameter "raid=partitionable" (or "raid=part") means
ian@0 52 that all auto-detected arrays are assembled as partitionable.
ian@0 53
ian@0 54 Boot time assembly of degraded/dirty arrays
ian@0 55 -------------------------------------------
ian@0 56
ian@0 57 If a raid5 or raid6 array is both dirty and degraded, it could have
ian@0 58 undetectable data corruption. This is because the fact that it is
ian@0 59 'dirty' means that the parity cannot be trusted, and the fact that it
ian@0 60 is degraded means that some datablocks are missing and cannot reliably
ian@0 61 be reconstructed (due to no parity).
ian@0 62
ian@0 63 For this reason, md will normally refuse to start such an array. This
ian@0 64 requires the sysadmin to take action to explicitly start the array
ian@0 65 desipite possible corruption. This is normally done with
ian@0 66 mdadm --assemble --force ....
ian@0 67
ian@0 68 This option is not really available if the array has the root
ian@0 69 filesystem on it. In order to support this booting from such an
ian@0 70 array, md supports a module parameter "start_dirty_degraded" which,
ian@0 71 when set to 1, bypassed the checks and will allows dirty degraded
ian@0 72 arrays to be started.
ian@0 73
ian@0 74 So, to boot with a root filesystem of a dirty degraded raid[56], use
ian@0 75
ian@0 76 md-mod.start_dirty_degraded=1
ian@0 77
ian@0 78
ian@0 79 Superblock formats
ian@0 80 ------------------
ian@0 81
ian@0 82 The md driver can support a variety of different superblock formats.
ian@0 83 Currently, it supports superblock formats "0.90.0" and the "md-1" format
ian@0 84 introduced in the 2.5 development series.
ian@0 85
ian@0 86 The kernel will autodetect which format superblock is being used.
ian@0 87
ian@0 88 Superblock format '0' is treated differently to others for legacy
ian@0 89 reasons - it is the original superblock format.
ian@0 90
ian@0 91
ian@0 92 General Rules - apply for all superblock formats
ian@0 93 ------------------------------------------------
ian@0 94
ian@0 95 An array is 'created' by writing appropriate superblocks to all
ian@0 96 devices.
ian@0 97
ian@0 98 It is 'assembled' by associating each of these devices with an
ian@0 99 particular md virtual device. Once it is completely assembled, it can
ian@0 100 be accessed.
ian@0 101
ian@0 102 An array should be created by a user-space tool. This will write
ian@0 103 superblocks to all devices. It will usually mark the array as
ian@0 104 'unclean', or with some devices missing so that the kernel md driver
ian@0 105 can create appropriate redundancy (copying in raid1, parity
ian@0 106 calculation in raid4/5).
ian@0 107
ian@0 108 When an array is assembled, it is first initialized with the
ian@0 109 SET_ARRAY_INFO ioctl. This contains, in particular, a major and minor
ian@0 110 version number. The major version number selects which superblock
ian@0 111 format is to be used. The minor number might be used to tune handling
ian@0 112 of the format, such as suggesting where on each device to look for the
ian@0 113 superblock.
ian@0 114
ian@0 115 Then each device is added using the ADD_NEW_DISK ioctl. This
ian@0 116 provides, in particular, a major and minor number identifying the
ian@0 117 device to add.
ian@0 118
ian@0 119 The array is started with the RUN_ARRAY ioctl.
ian@0 120
ian@0 121 Once started, new devices can be added. They should have an
ian@0 122 appropriate superblock written to them, and then passed be in with
ian@0 123 ADD_NEW_DISK.
ian@0 124
ian@0 125 Devices that have failed or are not yet active can be detached from an
ian@0 126 array using HOT_REMOVE_DISK.
ian@0 127
ian@0 128
ian@0 129 Specific Rules that apply to format-0 super block arrays, and
ian@0 130 arrays with no superblock (non-persistent).
ian@0 131 -------------------------------------------------------------
ian@0 132
ian@0 133 An array can be 'created' by describing the array (level, chunksize
ian@0 134 etc) in a SET_ARRAY_INFO ioctl. This must has major_version==0 and
ian@0 135 raid_disks != 0.
ian@0 136
ian@0 137 Then uninitialized devices can be added with ADD_NEW_DISK. The
ian@0 138 structure passed to ADD_NEW_DISK must specify the state of the device
ian@0 139 and it's role in the array.
ian@0 140
ian@0 141 Once started with RUN_ARRAY, uninitialized spares can be added with
ian@0 142 HOT_ADD_DISK.
ian@0 143
ian@0 144
ian@0 145
ian@0 146 MD devices in sysfs
ian@0 147 -------------------
ian@0 148 md devices appear in sysfs (/sys) as regular block devices,
ian@0 149 e.g.
ian@0 150 /sys/block/md0
ian@0 151
ian@0 152 Each 'md' device will contain a subdirectory called 'md' which
ian@0 153 contains further md-specific information about the device.
ian@0 154
ian@0 155 All md devices contain:
ian@0 156 level
ian@0 157 a text file indicating the 'raid level'. This may be a standard
ian@0 158 numerical level prefixed by "RAID-" - e.g. "RAID-5", or some
ian@0 159 other name such as "linear" or "multipath".
ian@0 160 If no raid level has been set yet (array is still being
ian@0 161 assembled), this file will be empty.
ian@0 162
ian@0 163 raid_disks
ian@0 164 a text file with a simple number indicating the number of devices
ian@0 165 in a fully functional array. If this is not yet known, the file
ian@0 166 will be empty. If an array is being resized (not currently
ian@0 167 possible) this will contain the larger of the old and new sizes.
ian@0 168 Some raid level (RAID1) allow this value to be set while the
ian@0 169 array is active. This will reconfigure the array. Otherwise
ian@0 170 it can only be set while assembling an array.
ian@0 171
ian@0 172 chunk_size
ian@0 173 This is the size if bytes for 'chunks' and is only relevant to
ian@0 174 raid levels that involve striping (1,4,5,6,10). The address space
ian@0 175 of the array is conceptually divided into chunks and consecutive
ian@0 176 chunks are striped onto neighbouring devices.
ian@0 177 The size should be atleast PAGE_SIZE (4k) and should be a power
ian@0 178 of 2. This can only be set while assembling an array
ian@0 179
ian@0 180 component_size
ian@0 181 For arrays with data redundancy (i.e. not raid0, linear, faulty,
ian@0 182 multipath), all components must be the same size - or at least
ian@0 183 there must a size that they all provide space for. This is a key
ian@0 184 part or the geometry of the array. It is measured in sectors
ian@0 185 and can be read from here. Writing to this value may resize
ian@0 186 the array if the personality supports it (raid1, raid5, raid6),
ian@0 187 and if the component drives are large enough.
ian@0 188
ian@0 189 metadata_version
ian@0 190 This indicates the format that is being used to record metadata
ian@0 191 about the array. It can be 0.90 (traditional format), 1.0, 1.1,
ian@0 192 1.2 (newer format in varying locations) or "none" indicating that
ian@0 193 the kernel isn't managing metadata at all.
ian@0 194
ian@0 195 level
ian@0 196 The raid 'level' for this array. The name will often (but not
ian@0 197 always) be the same as the name of the module that implements the
ian@0 198 level. To be auto-loaded the module must have an alias
ian@0 199 md-$LEVEL e.g. md-raid5
ian@0 200 This can be written only while the array is being assembled, not
ian@0 201 after it is started.
ian@0 202
ian@0 203 layout
ian@0 204 The "layout" for the array for the particular level. This is
ian@0 205 simply a number that is interpretted differently by different
ian@0 206 levels. It can be written while assembling an array.
ian@0 207
ian@0 208 resync_start
ian@0 209 The point at which resync should start. If no resync is needed,
ian@0 210 this will be a very large number. At array creation it will
ian@0 211 default to 0, though starting the array as 'clean' will
ian@0 212 set it much larger.
ian@0 213
ian@0 214 new_dev
ian@0 215 This file can be written but not read. The value written should
ian@0 216 be a block device number as major:minor. e.g. 8:0
ian@0 217 This will cause that device to be attached to the array, if it is
ian@0 218 available. It will then appear at md/dev-XXX (depending on the
ian@0 219 name of the device) and further configuration is then possible.
ian@0 220
ian@0 221 safe_mode_delay
ian@0 222 When an md array has seen no write requests for a certain period
ian@0 223 of time, it will be marked as 'clean'. When another write
ian@0 224 request arrive, the array is marked as 'dirty' before the write
ian@0 225 commenses. This is known as 'safe_mode'.
ian@0 226 The 'certain period' is controlled by this file which stores the
ian@0 227 period as a number of seconds. The default is 200msec (0.200).
ian@0 228 Writing a value of 0 disables safemode.
ian@0 229
ian@0 230 array_state
ian@0 231 This file contains a single word which describes the current
ian@0 232 state of the array. In many cases, the state can be set by
ian@0 233 writing the word for the desired state, however some states
ian@0 234 cannot be explicitly set, and some transitions are not allowed.
ian@0 235
ian@0 236 clear
ian@0 237 No devices, no size, no level
ian@0 238 Writing is equivalent to STOP_ARRAY ioctl
ian@0 239 inactive
ian@0 240 May have some settings, but array is not active
ian@0 241 all IO results in error
ian@0 242 When written, doesn't tear down array, but just stops it
ian@0 243 suspended (not supported yet)
ian@0 244 All IO requests will block. The array can be reconfigured.
ian@0 245 Writing this, if accepted, will block until array is quiessent
ian@0 246 readonly
ian@0 247 no resync can happen. no superblocks get written.
ian@0 248 write requests fail
ian@0 249 read-auto
ian@0 250 like readonly, but behaves like 'clean' on a write request.
ian@0 251
ian@0 252 clean - no pending writes, but otherwise active.
ian@0 253 When written to inactive array, starts without resync
ian@0 254 If a write request arrives then
ian@0 255 if metadata is known, mark 'dirty' and switch to 'active'.
ian@0 256 if not known, block and switch to write-pending
ian@0 257 If written to an active array that has pending writes, then fails.
ian@0 258 active
ian@0 259 fully active: IO and resync can be happening.
ian@0 260 When written to inactive array, starts with resync
ian@0 261
ian@0 262 write-pending
ian@0 263 clean, but writes are blocked waiting for 'active' to be written.
ian@0 264
ian@0 265 active-idle
ian@0 266 like active, but no writes have been seen for a while (safe_mode_delay).
ian@0 267
ian@0 268
ian@0 269 sync_speed_min
ian@0 270 sync_speed_max
ian@0 271 This are similar to /proc/sys/dev/raid/speed_limit_{min,max}
ian@0 272 however they only apply to the particular array.
ian@0 273 If no value has been written to these, of if the word 'system'
ian@0 274 is written, then the system-wide value is used. If a value,
ian@0 275 in kibibytes-per-second is written, then it is used.
ian@0 276 When the files are read, they show the currently active value
ian@0 277 followed by "(local)" or "(system)" depending on whether it is
ian@0 278 a locally set or system-wide value.
ian@0 279
ian@0 280 sync_completed
ian@0 281 This shows the number of sectors that have been completed of
ian@0 282 whatever the current sync_action is, followed by the number of
ian@0 283 sectors in total that could need to be processed. The two
ian@0 284 numbers are separated by a '/' thus effectively showing one
ian@0 285 value, a fraction of the process that is complete.
ian@0 286
ian@0 287 sync_speed
ian@0 288 This shows the current actual speed, in K/sec, of the current
ian@0 289 sync_action. It is averaged over the last 30 seconds.
ian@0 290
ian@0 291
ian@0 292 As component devices are added to an md array, they appear in the 'md'
ian@0 293 directory as new directories named
ian@0 294 dev-XXX
ian@0 295 where XXX is a name that the kernel knows for the device, e.g. hdb1.
ian@0 296 Each directory contains:
ian@0 297
ian@0 298 block
ian@0 299 a symlink to the block device in /sys/block, e.g.
ian@0 300 /sys/block/md0/md/dev-hdb1/block -> ../../../../block/hdb/hdb1
ian@0 301
ian@0 302 super
ian@0 303 A file containing an image of the superblock read from, or
ian@0 304 written to, that device.
ian@0 305
ian@0 306 state
ian@0 307 A file recording the current state of the device in the array
ian@0 308 which can be a comma separated list of
ian@0 309 faulty - device has been kicked from active use due to
ian@0 310 a detected fault
ian@0 311 in_sync - device is a fully in-sync member of the array
ian@0 312 writemostly - device will only be subject to read
ian@0 313 requests if there are no other options.
ian@0 314 This applies only to raid1 arrays.
ian@0 315 spare - device is working, but not a full member.
ian@0 316 This includes spares that are in the process
ian@0 317 of being recoverred to
ian@0 318 This list make grow in future.
ian@0 319 This can be written to.
ian@0 320 Writing "faulty" simulates a failure on the device.
ian@0 321 Writing "remove" removes the device from the array.
ian@0 322 Writing "writemostly" sets the writemostly flag.
ian@0 323 Writing "-writemostly" clears the writemostly flag.
ian@0 324
ian@0 325 errors
ian@0 326 An approximate count of read errors that have been detected on
ian@0 327 this device but have not caused the device to be evicted from
ian@0 328 the array (either because they were corrected or because they
ian@0 329 happened while the array was read-only). When using version-1
ian@0 330 metadata, this value persists across restarts of the array.
ian@0 331
ian@0 332 This value can be written while assembling an array thus
ian@0 333 providing an ongoing count for arrays with metadata managed by
ian@0 334 userspace.
ian@0 335
ian@0 336 slot
ian@0 337 This gives the role that the device has in the array. It will
ian@0 338 either be 'none' if the device is not active in the array
ian@0 339 (i.e. is a spare or has failed) or an integer less than the
ian@0 340 'raid_disks' number for the array indicating which possition
ian@0 341 it currently fills. This can only be set while assembling an
ian@0 342 array. A device for which this is set is assumed to be working.
ian@0 343
ian@0 344 offset
ian@0 345 This gives the location in the device (in sectors from the
ian@0 346 start) where data from the array will be stored. Any part of
ian@0 347 the device before this offset us not touched, unless it is
ian@0 348 used for storing metadata (Formats 1.1 and 1.2).
ian@0 349
ian@0 350 size
ian@0 351 The amount of the device, after the offset, that can be used
ian@0 352 for storage of data. This will normally be the same as the
ian@0 353 component_size. This can be written while assembling an
ian@0 354 array. If a value less than the current component_size is
ian@0 355 written, component_size will be reduced to this value.
ian@0 356
ian@0 357
ian@0 358 An active md device will also contain and entry for each active device
ian@0 359 in the array. These are named
ian@0 360
ian@0 361 rdNN
ian@0 362
ian@0 363 where 'NN' is the possition in the array, starting from 0.
ian@0 364 So for a 3 drive array there will be rd0, rd1, rd2.
ian@0 365 These are symbolic links to the appropriate 'dev-XXX' entry.
ian@0 366 Thus, for example,
ian@0 367 cat /sys/block/md*/md/rd*/state
ian@0 368 will show 'in_sync' on every line.
ian@0 369
ian@0 370
ian@0 371
ian@0 372 Active md devices for levels that support data redundancy (1,4,5,6)
ian@0 373 also have
ian@0 374
ian@0 375 sync_action
ian@0 376 a text file that can be used to monitor and control the rebuild
ian@0 377 process. It contains one word which can be one of:
ian@0 378 resync - redundancy is being recalculated after unclean
ian@0 379 shutdown or creation
ian@0 380 recover - a hot spare is being built to replace a
ian@0 381 failed/missing device
ian@0 382 idle - nothing is happening
ian@0 383 check - A full check of redundancy was requested and is
ian@0 384 happening. This reads all block and checks
ian@0 385 them. A repair may also happen for some raid
ian@0 386 levels.
ian@0 387 repair - A full check and repair is happening. This is
ian@0 388 similar to 'resync', but was requested by the
ian@0 389 user, and the write-intent bitmap is NOT used to
ian@0 390 optimise the process.
ian@0 391
ian@0 392 This file is writable, and each of the strings that could be
ian@0 393 read are meaningful for writing.
ian@0 394
ian@0 395 'idle' will stop an active resync/recovery etc. There is no
ian@0 396 guarantee that another resync/recovery may not be automatically
ian@0 397 started again, though some event will be needed to trigger
ian@0 398 this.
ian@0 399 'resync' or 'recovery' can be used to restart the
ian@0 400 corresponding operation if it was stopped with 'idle'.
ian@0 401 'check' and 'repair' will start the appropriate process
ian@0 402 providing the current state is 'idle'.
ian@0 403
ian@0 404 mismatch_count
ian@0 405 When performing 'check' and 'repair', and possibly when
ian@0 406 performing 'resync', md will count the number of errors that are
ian@0 407 found. The count in 'mismatch_cnt' is the number of sectors
ian@0 408 that were re-written, or (for 'check') would have been
ian@0 409 re-written. As most raid levels work in units of pages rather
ian@0 410 than sectors, this my be larger than the number of actual errors
ian@0 411 by a factor of the number of sectors in a page.
ian@0 412
ian@0 413 Each active md device may also have attributes specific to the
ian@0 414 personality module that manages it.
ian@0 415 These are specific to the implementation of the module and could
ian@0 416 change substantially if the implementation changes.
ian@0 417
ian@0 418 These currently include
ian@0 419
ian@0 420 stripe_cache_size (currently raid5 only)
ian@0 421 number of entries in the stripe cache. This is writable, but
ian@0 422 there are upper and lower limits (32768, 16). Default is 128.
ian@0 423 strip_cache_active (currently raid5 only)
ian@0 424 number of active entries in the stripe cache