Commit | Line | Data |
---|---|---|
c0a2fa1e | 1 | dm-raid |
be83651f | 2 | ======= |
9d09e663 | 3 | |
c0a2fa1e JB |
4 | The device-mapper RAID (dm-raid) target provides a bridge from DM to MD. |
5 | It allows the MD RAID drivers to be accessed using a device-mapper | |
6 | interface. | |
9d09e663 | 7 | |
be83651f JB |
8 | |
9 | Mapping Table Interface | |
10 | ----------------------- | |
c0a2fa1e JB |
11 | The target is named "raid" and it accepts the following parameters: |
12 | ||
13 | <raid_type> <#raid_params> <raid_params> \ | |
14 | <#raid_devs> <metadata_dev0> <dev0> [.. <metadata_devN> <devN>] | |
15 | ||
16 | <raid_type>: | |
b12d437b | 17 | raid1 RAID1 mirroring |
c0a2fa1e JB |
18 | raid4 RAID4 dedicated parity disk |
19 | raid5_la RAID5 left asymmetric | |
20 | - rotating parity 0 with data continuation | |
21 | raid5_ra RAID5 right asymmetric | |
22 | - rotating parity N with data continuation | |
23 | raid5_ls RAID5 left symmetric | |
24 | - rotating parity 0 with data restart | |
25 | raid5_rs RAID5 right symmetric | |
26 | - rotating parity N with data restart | |
27 | raid6_zr RAID6 zero restart | |
28 | - rotating parity zero (left-to-right) with data restart | |
29 | raid6_nr RAID6 N restart | |
30 | - rotating parity N (right-to-left) with data restart | |
31 | raid6_nc RAID6 N continue | |
32 | - rotating parity N (right-to-left) with data continuation | |
63f33b8d JB |
33 | raid10 Various RAID10 inspired algorithms chosen by additional params |
34 | - RAID10: Striped Mirrors (aka 'Striping on top of mirrors') | |
35 | - RAID1E: Integrated Adjacent Stripe Mirroring | |
fe5d2f4a | 36 | - RAID1E: Integrated Offset Stripe Mirroring |
63f33b8d | 37 | - and other similar RAID10 variants |
c0a2fa1e | 38 | |
40e47125 | 39 | Reference: Chapter 4 of |
c0a2fa1e JB |
40 | http://www.snia.org/sites/default/files/SNIA_DDF_Technical_Position_v2.0.pdf |
41 | ||
42 | <#raid_params>: The number of parameters that follow. | |
43 | ||
44 | <raid_params> consists of | |
45 | Mandatory parameters: | |
46 | <chunk_size>: Chunk size in sectors. This parameter is often known as | |
47 | "stripe size". It is the only mandatory parameter and | |
48 | is placed first. | |
49 | ||
50 | followed by optional parameters (in any order): | |
51 | [sync|nosync] Force or prevent RAID initialization. | |
52 | ||
be83651f | 53 | [rebuild <idx>] Rebuild drive number 'idx' (first drive is 0). |
c0a2fa1e JB |
54 | |
55 | [daemon_sleep <ms>] | |
56 | Interval between runs of the bitmap daemon that | |
57 | clear bits. A longer interval means less bitmap I/O but | |
58 | resyncing after a failure is likely to take longer. | |
59 | ||
60 | [min_recovery_rate <kB/sec/disk>] Throttle RAID initialization | |
61 | [max_recovery_rate <kB/sec/disk>] Throttle RAID initialization | |
be83651f JB |
62 | [write_mostly <idx>] Mark drive index 'idx' write-mostly. |
63 | [max_write_behind <sectors>] See '--write-behind=' (man mdadm) | |
64 | [stripe_cache <sectors>] Stripe cache size (RAID 4/5/6 only) | |
c1084561 JB |
65 | [region_size <sectors>] |
66 | The region_size multiplied by the number of regions is the | |
67 | logical size of the array. The bitmap records the device | |
68 | synchronisation state for each region. | |
c0a2fa1e | 69 | |
63f33b8d | 70 | [raid10_copies <# copies>] |
fe5d2f4a | 71 | [raid10_format <near|far|offset>] |
63f33b8d JB |
72 | These two options are used to alter the default layout of |
73 | a RAID10 configuration. The number of copies is can be | |
fe5d2f4a JB |
74 | specified, but the default is 2. There are also three |
75 | variations to how the copies are laid down - the default | |
76 | is "near". Near copies are what most people think of with | |
77 | respect to mirroring. If these options are left unspecified, | |
78 | or 'raid10_copies 2' and/or 'raid10_format near' are given, | |
79 | then the layouts for 2, 3 and 4 devices are: | |
63f33b8d JB |
80 | 2 drives 3 drives 4 drives |
81 | -------- ---------- -------------- | |
82 | A1 A1 A1 A1 A2 A1 A1 A2 A2 | |
83 | A2 A2 A2 A3 A3 A3 A3 A4 A4 | |
84 | A3 A3 A4 A4 A5 A5 A5 A6 A6 | |
85 | A4 A4 A5 A6 A6 A7 A7 A8 A8 | |
86 | .. .. .. .. .. .. .. .. .. | |
87 | The 2-device layout is equivalent 2-way RAID1. The 4-device | |
88 | layout is what a traditional RAID10 would look like. The | |
89 | 3-device layout is what might be called a 'RAID1E - Integrated | |
90 | Adjacent Stripe Mirroring'. | |
91 | ||
fe5d2f4a JB |
92 | If 'raid10_copies 2' and 'raid10_format far', then the layouts |
93 | for 2, 3 and 4 devices are: | |
94 | 2 drives 3 drives 4 drives | |
95 | -------- -------------- -------------------- | |
96 | A1 A2 A1 A2 A3 A1 A2 A3 A4 | |
97 | A3 A4 A4 A5 A6 A5 A6 A7 A8 | |
98 | A5 A6 A7 A8 A9 A9 A10 A11 A12 | |
99 | .. .. .. .. .. .. .. .. .. | |
100 | A2 A1 A3 A1 A2 A2 A1 A4 A3 | |
101 | A4 A3 A6 A4 A5 A6 A5 A8 A7 | |
102 | A6 A5 A9 A7 A8 A10 A9 A12 A11 | |
103 | .. .. .. .. .. .. .. .. .. | |
104 | ||
105 | If 'raid10_copies 2' and 'raid10_format offset', then the | |
106 | layouts for 2, 3 and 4 devices are: | |
107 | 2 drives 3 drives 4 drives | |
108 | -------- ------------ ----------------- | |
109 | A1 A2 A1 A2 A3 A1 A2 A3 A4 | |
110 | A2 A1 A3 A1 A2 A2 A1 A4 A3 | |
111 | A3 A4 A4 A5 A6 A5 A6 A7 A8 | |
112 | A4 A3 A6 A4 A5 A6 A5 A8 A7 | |
113 | A5 A6 A7 A8 A9 A9 A10 A11 A12 | |
114 | A6 A5 A9 A7 A8 A10 A9 A12 A11 | |
115 | .. .. .. .. .. .. .. .. .. | |
116 | Here we see layouts closely akin to 'RAID1E - Integrated | |
117 | Offset Stripe Mirroring'. | |
118 | ||
c0a2fa1e JB |
119 | <#raid_devs>: The number of devices composing the array. |
120 | Each device consists of two entries. The first is the device | |
121 | containing the metadata (if any); the second is the one containing the | |
b12d437b | 122 | data. |
c0a2fa1e JB |
123 | |
124 | If a drive has failed or is missing at creation time, a '-' can be | |
125 | given for both the metadata and data drives for a given position. | |
126 | ||
127 | ||
be83651f | 128 | Example Tables |
c0a2fa1e | 129 | -------------- |
b12d437b | 130 | # RAID4 - 4 data drives, 1 parity (no metadata devices) |
9d09e663 N |
131 | # No metadata devices specified to hold superblock/bitmap info |
132 | # Chunk size of 1MiB | |
133 | # (Lines separated for easy reading) | |
c0a2fa1e | 134 | |
9d09e663 N |
135 | 0 1960893648 raid \ |
136 | raid4 1 2048 \ | |
137 | 5 - 8:17 - 8:33 - 8:49 - 8:65 - 8:81 | |
138 | ||
b12d437b | 139 | # RAID4 - 4 data drives, 1 parity (with metadata devices) |
9d09e663 N |
140 | # Chunk size of 1MiB, force RAID initialization, |
141 | # min recovery rate at 20 kiB/sec/disk | |
c0a2fa1e | 142 | |
9d09e663 | 143 | 0 1960893648 raid \ |
b12d437b JB |
144 | raid4 4 2048 sync min_recovery_rate 20 \ |
145 | 5 8:17 8:18 8:33 8:34 8:49 8:50 8:65 8:66 8:81 8:82 | |
9d09e663 | 146 | |
be83651f JB |
147 | |
148 | Status Output | |
149 | ------------- | |
c0a2fa1e | 150 | 'dmsetup table' displays the table used to construct the mapping. |
46bed2b5 | 151 | The optional parameters are always printed in the order listed |
c0a2fa1e JB |
152 | above with "sync" or "nosync" always output ahead of the other |
153 | arguments, regardless of the order used when originally loading the table. | |
46bed2b5 | 154 | Arguments that can be repeated are ordered by value. |
9d09e663 | 155 | |
be83651f JB |
156 | |
157 | 'dmsetup status' yields information on the state and health of the array. | |
158 | The output is as follows (normally a single line, but expanded here for | |
159 | clarity): | |
9d09e663 | 160 | 1: <s> <l> raid \ |
be83651f JB |
161 | 2: <raid_type> <#devices> <health_chars> \ |
162 | 3: <sync_ratio> <sync_action> <mismatch_cnt> | |
9d09e663 | 163 | |
c0a2fa1e | 164 | Line 1 is the standard output produced by device-mapper. |
be83651f JB |
165 | Line 2 & 3 are produced by the raid target and are best explained by example: |
166 | 0 1960893648 raid raid4 5 AAAAA 2/490221568 init 0 | |
9d09e663 | 167 | Here we can see the RAID type is raid4, there are 5 devices - all of |
be83651f JB |
168 | which are 'A'live, and the array is 2/490221568 complete with its initial |
169 | recovery. Here is a fuller description of the individual fields: | |
170 | <raid_type> Same as the <raid_type> used to create the array. | |
171 | <health_chars> One char for each device, indicating: 'A' = alive and | |
172 | in-sync, 'a' = alive but not in-sync, 'D' = dead/failed. | |
173 | <sync_ratio> The ratio indicating how much of the array has undergone | |
174 | the process described by 'sync_action'. If the | |
175 | 'sync_action' is "check" or "repair", then the process | |
176 | of "resync" or "recover" can be considered complete. | |
177 | <sync_action> One of the following possible states: | |
178 | idle - No synchronization action is being performed. | |
179 | frozen - The current action has been halted. | |
180 | resync - Array is undergoing its initial synchronization | |
181 | or is resynchronizing after an unclean shutdown | |
182 | (possibly aided by a bitmap). | |
183 | recover - A device in the array is being rebuilt or | |
184 | replaced. | |
185 | check - A user-initiated full check of the array is | |
186 | being performed. All blocks are read and | |
187 | checked for consistency. The number of | |
188 | discrepancies found are recorded in | |
189 | <mismatch_cnt>. No changes are made to the | |
190 | array by this action. | |
191 | repair - The same as "check", but discrepancies are | |
192 | corrected. | |
193 | reshape - The array is undergoing a reshape. | |
194 | <mismatch_cnt> The number of discrepancies found between mirror copies | |
195 | in RAID1/10 or wrong parity values found in RAID4/5/6. | |
196 | This value is valid only after a "check" of the array | |
197 | is performed. A healthy array has a 'mismatch_cnt' of 0. | |
198 | ||
199 | Message Interface | |
200 | ----------------- | |
201 | The dm-raid target will accept certain actions through the 'message' interface. | |
202 | ('man dmsetup' for more information on the message interface.) These actions | |
203 | include: | |
204 | "idle" - Halt the current sync action. | |
205 | "frozen" - Freeze the current sync action. | |
206 | "resync" - Initiate/continue a resync. | |
207 | "recover"- Initiate/continue a recover process. | |
208 | "check" - Initiate a check (i.e. a "scrub") of the array. | |
209 | "repair" - Initiate a repair of the array. | |
210 | "reshape"- Currently unsupported (-EINVAL). | |
4ec1e369 JB |
211 | |
212 | Version History | |
213 | --------------- | |
214 | 1.0.0 Initial version. Support for RAID 4/5/6 | |
215 | 1.1.0 Added support for RAID 1 | |
216 | 1.2.0 Handle creation of arrays that contain failed devices. | |
217 | 1.3.0 Added support for RAID 10 | |
218 | 1.3.1 Allow device replacement/rebuild for RAID 10 | |
55ebbb59 | 219 | 1.3.2 Fix/improve redundancy checking for RAID10 |
fe5d2f4a | 220 | 1.4.0 Non-functional change. Removes arg from mapping function. |
be83651f JB |
221 | 1.4.1 RAID10 fix redundancy validation checks (commit 55ebbb5). |
222 | 1.4.2 Add RAID10 "far" and "offset" algorithm support. | |
223 | 1.5.0 Add message interface to allow manipulation of the sync_action. | |
224 | New status (STATUSTYPE_INFO) fields: sync_action and mismatch_cnt. | |
9092c02d | 225 | 1.5.1 Add ability to restore transiently failed devices on resume. |
c4a39551 | 226 | 1.5.2 'mismatch_cnt' is zero unless [last_]sync_action is "check". |