1 =======
2 dm-raid
3 =======
4
5 The device-mapper RAID (dm-raid) target provid
6 It allows the MD RAID drivers to be accessed u
7 interface.
8
9
10 Mapping Table Interface
11 -----------------------
12 The target is named "raid" and it accepts the
13
14 <raid_type> <#raid_params> <raid_params> \
15 <#raid_devs> <metadata_dev0> <dev0> [.. <m
16
17 <raid_type>:
18
19 ============= ==============================
20 raid0 RAID0 striping (no resilience)
21 raid1 RAID1 mirroring
22 raid4 RAID4 with dedicated last pari
23 raid5_n RAID5 with dedicated last pari
24 Same as raid4
25
26 - Transitory layout
27 raid5_la RAID5 left asymmetric
28
29 - rotating parity 0 with data
30 raid5_ra RAID5 right asymmetric
31
32 - rotating parity N with data
33 raid5_ls RAID5 left symmetric
34
35 - rotating parity 0 with data
36 raid5_rs RAID5 right symmetric
37
38 - rotating parity N with data
39 raid6_zr RAID6 zero restart
40
41 - rotating parity zero (left-t
42 raid6_nr RAID6 N restart
43
44 - rotating parity N (right-to-
45 raid6_nc RAID6 N continue
46
47 - rotating parity N (right-to-
48 raid6_n_6 RAID6 with dedicate parity dis
49
50 - parity and Q-syndrome on the
51 layout for takeover from/to
52 raid6_la_6 Same as "raid_la" plus dedicat
53
54 - layout for takeover from rai
55 raid6_ra_6 Same as "raid5_ra" dedicated l
56
57 - layout for takeover from rai
58 raid6_ls_6 Same as "raid5_ls" dedicated l
59
60 - layout for takeover from rai
61 raid6_rs_6 Same as "raid5_rs" dedicated l
62
63 - layout for takeover from rai
64 raid10 Various RAID10 inspired algori
65 (see raid10_format and raid10_
66
67 - RAID10: Striped Mirrors (aka
68 - RAID1E: Integrated Adjacent
69 - RAID1E: Integrated Offset St
70 - and other similar RAID10 var
71 ============= ==============================
72
73 Reference: Chapter 4 of
74 https://www.snia.org/sites/default/files/SNI
75
76 <#raid_params>: The number of parameters that
77
78 <raid_params> consists of
79
80 Mandatory parameters:
81 <chunk_size>:
82 Chunk size in sectors.
83 "stripe size". It is th
84 is placed first.
85
86 followed by optional parameters (in any or
87 [sync|nosync]
88 Force or prevent RAID initiali
89
90 [rebuild <idx>]
91 Rebuild drive number 'idx' (fi
92
93 [daemon_sleep <ms>]
94 Interval between runs of the b
95 clear bits. A longer interval
96 resyncing after a failure is l
97
98 [min_recovery_rate <kB/sec/disk>]
99 Throttle RAID initialization
100 [max_recovery_rate <kB/sec/disk>]
101 Throttle RAID initialization
102 [write_mostly <idx>]
103 Mark drive index 'idx' write-m
104 [max_write_behind <sectors>]
105 See '--write-behind=' (man mda
106 [stripe_cache <sectors>]
107 Stripe cache size (RAID 4/5/6
108 [region_size <sectors>]
109 The region_size multiplied by
110 logical size of the array. Th
111 synchronisation state for each
112
113 [raid10_copies <# copies>], [raid10_
114 These two options are used to
115 a RAID10 configuration. The n
116 specified, but the default is
117 variations to how the copies a
118 is "near". Near copies are wh
119 respect to mirroring. If thes
120 or 'raid10_copies 2' and/or 'r
121 then the layouts for 2, 3 and
122
123 ======== ==========
124 2 drives 3 drives
125 ======== ==========
126 A1 A1 A1 A1 A2
127 A2 A2 A2 A3 A3
128 A3 A3 A4 A4 A5
129 A4 A4 A5 A6 A6
130 .. .. .. .. ..
131 ======== ==========
132
133 The 2-device layout is equival
134 layout is what a traditional R
135 3-device layout is what might
136 Adjacent Stripe Mirroring'.
137
138 If 'raid10_copies 2' and 'raid
139 for 2, 3 and 4 devices are:
140
141 ======== =========
142 2 drives 3 drives
143 ======== =========
144 A1 A2 A1 A2
145 A3 A4 A4 A5
146 A5 A6 A7 A8
147 .. .. .. ..
148 A2 A1 A3 A1
149 A4 A3 A6 A4
150 A6 A5 A9 A7
151 .. .. .. ..
152 ======== =========
153
154 If 'raid10_copies 2' and 'raid
155 layouts for 2, 3 and 4 devices
156
157 ======== ==========
158 2 drives 3 drives
159 ======== ==========
160 A1 A2 A1 A2 A3
161 A2 A1 A3 A1 A2
162 A3 A4 A4 A5 A6
163 A4 A3 A6 A4 A5
164 A5 A6 A7 A8 A9
165 A6 A5 A9 A7 A8
166 .. .. .. .. ..
167 ======== ==========
168
169 Here we see layouts closely ak
170 Offset Stripe Mirroring'.
171
172 [delta_disks <N>]
173 The delta_disks option value (
174 device removal (negative value
175 value) to any reshape supporti
176 RAID levels 4/5/6 allow for ad
177 and data device tuple), raid10
178 allow for device addition. rai
179 reshaping at all.
180 A minimum of devices have to b
181 which is 3 devices for raid4/5
182
183 [data_offset <sectors>]
184 This option value defines the
185 where the data starts. This is
186 reshaping space to avoid writi
187 changing the layout of stripes
188 may happen at any time without
189 E.g. when adding devices to an
190 forward reshaping, the out-of-
191 at the beginning of each raid
192 MD personalities supporting su
193 the existing first stripes (th
194 starting at data_offset to fil
195 number of stripes, calculate t
196 and write that new stripe to o
197 N-1 other new stripes. This ou
198 the RAID type (i.e. the alloca
199 changing from raid5_ls to raid
200
201 [journal_dev <dev>]
202 This option adds a journal dev
203 uses it to close the 'write ho
204 to the component devices which
205 The journal device is used as
206 be throttled versus non-journa
207 Takeover/reshape is not possib
208 it has to be deconfigured befo
209
210 [journal_mode <mode>]
211 This option sets the caching m
212 (see 'journal_dev <dev>' above
213 If 'writeback' is selected the
214 and must not suffer from the '
215 raid1 or raid10) to avoid a si
216
217 <#raid_devs>: The number of devices composing
218 Each device consists of two entries.
219 containing the metadata (if any); the
220 data. A Maximum of 64 metadata/data de
221 up to target version 1.8.0.
222 1.9.0 supports up to 253 which is enfo
223
224 If a drive has failed or is missing at
225 given for both the metadata and data d
226
227
228 Example Tables
229 --------------
230
231 ::
232
233 # RAID4 - 4 data drives, 1 parity (no metada
234 # No metadata devices specified to hold supe
235 # Chunk size of 1MiB
236 # (Lines separated for easy reading)
237
238 0 1960893648 raid \
239 raid4 1 2048 \
240 5 - 8:17 - 8:33 - 8:49 - 8:65 - 8:81
241
242 # RAID4 - 4 data drives, 1 parity (with meta
243 # Chunk size of 1MiB, force RAID initializat
244 # min recovery rate at 20 kiB/sec/disk
245
246 0 1960893648 raid \
247 raid4 4 2048 sync min_recovery_rate
248 5 8:17 8:18 8:33 8:34 8:49 8:50 8:65
249
250
251 Status Output
252 -------------
253 'dmsetup table' displays the table used to con
254 The optional parameters are always printed in
255 above with "sync" or "nosync" always output ah
256 arguments, regardless of the order used when o
257 Arguments that can be repeated are ordered by
258
259
260 'dmsetup status' yields information on the sta
261 The output is as follows (normally a single li
262 clarity)::
263
264 1: <s> <l> raid \
265 2: <raid_type> <#devices> <health_chars
266 3: <sync_ratio> <sync_action> <mismatch
267
268 Line 1 is the standard output produced by devi
269
270 Line 2 & 3 are produced by the raid target and
271
272 0 1960893648 raid raid4 5 AAAAA 2/4902
273
274 Here we can see the RAID type is raid4, there
275 which are 'A'live, and the array is 2/49022156
276 recovery. Here is a fuller description of the
277
278 =============== ======================
279 <raid_type> Same as the <raid_type
280 <health_chars> One char for each devi
281
282 - 'A' = alive and in-s
283 - 'a' = alive but not
284 - 'D' = dead/failed.
285 <sync_ratio> The ratio indicating h
286 the process described
287 'sync_action' is "chec
288 of "resync" or "recove
289 <sync_action> One of the following p
290
291 idle
292 - No synchroni
293 frozen
294 - The current
295 resync
296 - Array is und
297 or is resync
298 (possibly ai
299 recover
300 - A device in
301 replaced.
302 check
303 - A user-initi
304 being perfor
305 checked for
306 discrepancie
307 <mismatch_cn
308 array by thi
309 repair
310 - The same as
311 corrected.
312 reshape
313 - The array is
314 <mismatch_cnt> The number of discrepa
315 in RAID1/10 or wrong p
316 This value is valid on
317 is performed. A healt
318 <data_offset> The current data offse
319 each component device
320 raid parameter to supp
321 <journal_char> - 'A' - active write-t
322 - 'a' - active write-b
323 - 'D' - dead journal d
324 - '-' - no journal dev
325 =============== ======================
326
327
328 Message Interface
329 -----------------
330 The dm-raid target will accept certain actions
331 ('man dmsetup' for more information on the mes
332 include:
333
334 ========= ============================
335 "idle" Halt the current sync action
336 "frozen" Freeze the current sync acti
337 "resync" Initiate/continue a resync.
338 "recover" Initiate/continue a recover
339 "check" Initiate a check (i.e. a "sc
340 "repair" Initiate a repair of the arr
341 ========= ============================
342
343
344 Discard Support
345 ---------------
346 The implementation of discard support among ha
347 When a block is discarded, some storage device
348 the block is read. These devices set the 'dis
349 attribute. Other devices will return random d
350 devices that advertise 'discard_zeroes_data' w
351 zeroes when discarded blocks are read! Since
352 from a number of devices to calculate parity b
353 reasons) relies on 'discard_zeroes_data' being
354 that the devices be consistent. Blocks may be
355 of a RAID 4/5/6 stripe and if subsequent read
356 consistent, the parity blocks may be calculate
357 making the parity blocks useless for redundanc
358 understand how your hardware behaves with disc
359 enable discards with RAID 4/5/6.
360
361 Since the behavior of storage devices is unrel
362 even when reporting 'discard_zeroes_data', by
363 discard support is disabled -- this ensures da
364 expense of losing some performance.
365
366 Storage devices that properly support 'discard
367 increasingly whitelisted in the kernel and can
368
369 For trusted devices, the following dm-raid mod
370 to safely enable discard support for RAID 4/5/
371
372 'devices_handle_discards_safely'
373
374
375 Version History
376 ---------------
377
378 ::
379
380 1.0.0 Initial version. Support for RAID 4/5
381 1.1.0 Added support for RAID 1
382 1.2.0 Handle creation of arrays that contain
383 1.3.0 Added support for RAID 10
384 1.3.1 Allow device replacement/rebuild for R
385 1.3.2 Fix/improve redundancy checking for RA
386 1.4.0 Non-functional change. Removes arg fr
387 1.4.1 RAID10 fix redundancy validation check
388 1.4.2 Add RAID10 "far" and "offset" algorith
389 1.5.0 Add message interface to allow manipul
390 New status (STATUSTYPE_INFO) fields: s
391 1.5.1 Add ability to restore transiently fai
392 1.5.2 'mismatch_cnt' is zero unless [last_]s
393 1.6.0 Add discard support (and devices_handl
394 1.7.0 Add support for MD RAID0 mappings.
395 1.8.0 Explicitly check for compatible flags
396 and reject to start the raid set if an
397 target version, thus avoiding data cor
398 with a reshape in progress.
399 1.9.0 Add support for RAID level takeover/re
400 and set size reduction.
401 1.9.1 Fix activation of existing RAID 4/10 m
402 1.9.2 Don't emit '- -' on the status table l
403 fails reading a superblock. Correctly
404 'D' on the status line. If '- -' is p
405 '- -' on the table line and '-' as the
406 1.10.0 Add support for raid4/5/6 journal devi
407 1.10.1 Fix data corruption on reshape request
408 1.11.0 Fix table line argument order
409 (wrong raid10_copies/raid10_format seq
410 1.11.1 Add raid4/5/6 journal write-back suppo
411 1.12.1 Fix for MD deadlock between mddev_susp
412 1.13.0 Fix dev_health status at end of "recov
413 1.13.1 Fix deadlock caused by early md_stop_w
414 state races.
415 1.13.2 Fix raid redundancy validation and avo
416 1.14.0 Fix reshape race on small devices. Fi
417 deadlock/potential data corruption. U
418 specific devices are requested via reb
419 rebuild errors.
420 1.15.0 Fix size extensions not being synchron
421 pages allocated; also fix those not o
422 1.15.1 Fix argument count and arguments for r
423 on the status line.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.