3 md \- Multiple Device driver aka Linux Software Raid
11 driver provides virtual devices that are created from one or more
12 independent underlying devices. This array of devices often contains
13 redundancy, and hence the acronym RAID which stands for a Redundant
14 Array of Independent Devices.
17 supports RAID levels 1 (mirroring) 4 (striped array with parity
18 device), 5 (striped array with distributed parity information) and 6
19 (striped array with distributed dual redundancy information.) If
20 some number of underlying devices fails while using one of these
21 levels, the array will continue to function; this number is one for
22 RAID levels 4 and 5, two for RAID level 6, and all but one (N-1) for
26 also supports a number of pseudo RAID (non-redundant) configurations
27 including RAID0 (striped array), LINEAR (catenated array),
28 MULTIPATH (a set of different interfaces to the same device),
29 and FAULTY (a layer over a single device into which errors can be injected).
32 Each device in an array may have a
34 which records information about the structure and state of the array.
35 This allows the array to be reliably re-assembled after a shutdown.
37 From Linux kernel version 2.6.10,
39 provides support for two different formats of this superblock, and
40 other formats can be added. Prior to this release, only one format is
43 The common format - known as version 0.90 - has
44 a superblock that is 4K long and is written into a 64K aligned block that
45 starts at least 64K and less than 128K from the end of the device
46 (i.e. to get the address of the superblock round the size of the
47 device down to a multiple of 64K and then subtract 64K).
48 The available size of each device is the amount of space before the
49 super block, so between 64K and 128K is lost when a device in
50 incorporated into an MD array.
51 This superblock stores multi-byte fields in a processor-dependant
52 manner, so arrays cannot easily be moved between computers with
55 The new format - known as version 1 - has a superblock that is
56 normally 1K long, but can be longer. It is normally stored between 8K
57 and 12K from the end of the device, on a 4K boundary, though
58 variations can be stored at the start of the device (version 1.1) or 4K from
59 the start of the device (version 1.2).
60 This superblock format stores multibyte data in a
61 processor-independant format and has supports upto hundreds of
62 component devices (version 0.90 only supports 28).
64 The superblock contains, among other things:
67 The manner in which the devices are arranged into the array
68 (linear, raid0, raid1, raid4, raid5, multipath).
71 a 128 bit Universally Unique Identifier that identifies the array that
72 this device is part of.
74 .SS ARRAYS WITHOUT SUPERBLOCKS
75 While it is usually best to create arrays with superblocks so that
76 they can be assembled reliably, there are some circumstances where an
77 array without superblocks in preferred. This include:
82 driver only supported Linear and Raid0 configurations and did not use
83 a superblock (which is less critical with these configurations).
84 While such arrays should be rebuilt with superblocks if possible,
86 continues to support them.
89 Being a largely transparent layer over a different device, the FAULTY
90 personality doesn't gain anything from having a superblock.
93 It is often possible to detect devices which are different paths to
94 the same storage directly rather than having a distinctive superblock
95 written to the device and searched for on all paths. In this case,
96 a MULTIPATH array with no superblock makes sense.
99 In some configurations it might be desired to create a raid1
100 configuration that does use a superblock, and to maintain the state of
101 the array elsewhere. While not encouraged, this is supported.
105 A linear array simply catenates the available space on each
106 drive together to form one large virtual drive.
108 One advantage of this arrangement over the more common RAID0
109 arrangement is that the array may be reconfigured at a later time with
110 an extra drive and so the array is made bigger without disturbing the
111 data that is on the array. However this cannot be done on a live
117 A RAID0 array (which has zero redundancy) is also known as a
119 A RAID0 array is configured at creation with a
121 which must be a power of two, and at least 4 kibibytes.
123 The RAID0 driver assigns the first chunk of the array to the first
124 device, the second chunk to the second device, and so on until all
125 drives have been assigned one chunk. This collection of chunks forms
128 Further chunks are gathered into stripes in the same way which are
129 assigned to the remaining space in the drives.
131 If devices in the array are not all the same size, then once the
132 smallest device has been exhausted, the RAID0 driver starts
133 collecting chunks into smaller stripes that only span the drives which
134 still have remaining space.
139 A RAID1 array is also known as a mirrored set (though mirrors tend to
140 provide reflected images, which RAID1 does not) or a plex.
142 Once initialised, each device in a RAID1 array contains exactly the
143 same data. Changes are written to all devices in parallel. Data is
144 read from any one device. The driver attempts to distribute read
145 requests across all devices to maximise performance.
147 All devices in a RAID1 array should be the same size. If they are
148 not, then only the amount of space available on the smallest device is
149 used. Any extra space on other devices is wasted.
153 A RAID4 array is like a RAID0 array with an extra device for storing
154 parity. This device is the last of the active devices in the
155 array. Unlike RAID0, RAID4 also requires that all stripes span all
156 drives, so extra space on devices that are larger than the smallest is
159 When any block in a RAID4 array is modified the parity block for that
160 stripe (i.e. the block in the parity device at the same device offset
161 as the stripe) is also modified so that the parity block always
162 contains the "parity" for the whole stripe. i.e. its contents is
163 equivalent to the result of performing an exclusive-or operation
164 between all the data blocks in the stripe.
166 This allows the array to continue to function if one device fails.
167 The data that was on that device can be calculated as needed from the
168 parity block and the other data blocks.
172 RAID5 is very similar to RAID4. The difference is that the parity
173 blocks for each stripe, instead of being on a single device, are
174 distributed across all devices. This allows more parallelism when
175 writing as two different block updates will quite possibly affect
176 parity blocks on different devices so there is less contention.
178 This also allows more parallelism when reading as read requests are
179 distributed over all the devices in the array instead of all but one.
183 RAID6 is similar to RAID5, but can handle the loss of any \fItwo\fP
184 devices without data loss. Accordingly, it requires N+2 drives to
185 store N drives worth of data.
187 The performance for RAID6 is slightly lower but comparable to RAID5 in
188 normal mode and single disk failure mode. It is very slow in dual
189 disk failure mode, however.
193 MULTIPATH is not really a RAID at all as there is only one real device
194 in a MULTIPATH md array. However there are multiple access points
195 (paths) to this device, and one of these paths might fail, so there
196 are some similarities.
198 A MULTIPATH array is composed of a number of logical different
199 devices, often fibre channel interfaces, that all refer the the same
200 real device. If one of these interfaces fails (e.g. due to cable
201 problems), the multipath driver to attempt to redirect requests to
205 The FAULTY md module is provided for testing purposes. A faulty array
206 has exactly one component device and is normally assembled without a
207 superblock, so the md array created provides direct access to all of
208 the data in the component device.
210 The FAULTY module may be requested to simulate faults to allow testing
211 of other md levels or of filesystem. Faults can be chosen to trigger
212 on read requests or write requests, and can be transient (a subsequent
213 read/write at the address will probably succeed) or persistant
214 (subsequent read/write of the same address will fail). Further, read
215 faults can be "fixable" meaning that they persist until a write
216 request at the same address.
218 Fault types can be requested with a period. In this case the fault
219 will recur repeatedly after the given number of request of the
220 relevant time. For example if persistent read faults have a period of
221 100, then ever 100th read request would generate a fault, and the
222 faulty sector would be recorded so that subsequent reads on that
223 sector would also fail.
225 There is a limit to the number of faulty sectors that are remembered.
226 Faults generated after this limit is exhausted are treated as
229 It list of faulty sectors can be flushed, and the active list of
230 failure modes can be cleared.
234 When changes are made to a RAID1, RAID4, RAID5 or RAID6 array there is a
235 possibility of inconsistency for short periods of time as each update
236 requires are least two block to be written to different devices, and
237 these writes probably wont happen at exactly the same time.
238 Thus if a system with one of these arrays is shutdown in the middle of
239 a write operation (e.g. due to power failure), the array may not be
242 To handle this situation, the md driver marks an array as "dirty"
243 before writing any data to it, and marks it as "clean" when the array
244 is being disabled, e.g. at shutdown. If the md driver finds an array
245 to be dirty at startup, it proceeds to correct any possibly
246 inconsistency. For RAID1, this involves copying the contents of the
247 first drive onto all other drives. For RAID4, RAID5 and RAID6 this
248 involves recalculating the parity for each stripe and making sure that
249 the parity block has the correct data. This process, known as
250 "resynchronising" or "resync" is performed in the background. The
251 array can still be used, though possibly with reduced performance.
253 If a RAID4, RAID5 or RAID6 array is degraded (missing at least one
254 drive) when it is restarted after an unclean shutdown, it cannot
255 recalculate parity, and so it is possible that data might be
256 undetectably corrupted. The 2.4 md driver
258 alert the operator to this condition. The 2.5 md driver will fail to
259 start an array in this condition without manual intervention.
263 If the md driver detects any error on a device in a RAID1, RAID4,
264 RAID5 or RAID6 array, it immediately disables that device (marking it
265 as faulty) and continues operation on the remaining devices. If there
266 is a spare drive, the driver will start recreating on one of the spare
267 drives the data what was on that failed drive, either by copying a
268 working drive in a RAID1 configuration, or by doing calculations with
269 the parity block on RAID4, RAID5 or RAID6.
271 While this recovery process is happening, the md driver will monitor
272 accesses to the array and will slow down the rate of recovery if other
273 activity is happening, so that normal access to the array will not be
274 unduly affected. When no other activity is happening, the recovery
275 process proceeds at full speed. The actual speed targets for the two
276 different situations can be controlled by the
280 control files mentioned below.
282 .SS KERNEL PARAMETERS
284 The md driver recognised three different kernel parameters.
287 This will disable the normal detection of md arrays that happens at
288 boot time. If a drive is partitioned with MS-DOS style partitions,
289 then if any of the 4 main partitions has a partition type of 0xFD,
290 then that partition will normally be inspected to see if it is part of
291 an MD array, and if any full arrays are found, they are started. This
292 kernel paramenter disables this behaviour.
295 .BI md= n , dev , dev ,...
296 This tells the md driver to assemble
298 from the listed devices. It is only necessary to start the device
299 holding the root filesystem this way. Other arrays are best started
300 once the system is booted.
303 .BI md= n , l , c , i , dev...
304 This tells the md driver to assemble a legacy RAID0 or LINEAR array
305 without a superblock.
307 gives the md device number,
309 gives the level, 0 for RAID0 or -1 for LINEAR,
311 gives the chunk size as a base-2 logarithm offset by twelve, so 0
312 means 4K, 1 means 8K.
314 is ignored (legacy support).
319 Contains information about the status of currently running array.
321 .B /proc/sys/dev/raid/speed_limit_min
322 A readable and writable file that reflects the current goal rebuild
323 speed for times when non-rebuild activity is current on an array.
324 The speed is in Kibibytes per second, and is a per-device rate, not a
325 per-array rate (which means that an array with more disc will shuffle
326 more data for a given speed). The default is 100.
329 .B /proc/sys/dev/raid/speed_limit_max
330 A readable and writable file that reflects the current goal rebuild
331 speed for times when no non-rebuild activity is current on an array.
332 The default is 100,000.