1BTRFS-DEVICE(8) BTRFS BTRFS-DEVICE(8)
2
6 btrfs-device - manage devices of btrfs filesystems
7
9 btrfs device <subcommand> <args>
10
12 The btrfs device command group is used to manage devices of the btrfs
13 filesystems.
14
16 BTRFS filesystem can be created on top of single or multiple block de‐
17 vices. Devices can be then added, removed or replaced on demand. Data
18 and metadata are organized in allocation profiles with various redun‐
19 dancy policies. There's some similarity with traditional RAID levels,
20 but this could be confusing to users familiar with the traditional
21 meaning. Due to the similarity, the RAID terminology is widely used in
22 the documentation. See mkfs.btrfs(8) for more details and the exact
23 profile capabilities and constraints.
24 The device management works on a mounted filesystem. Devices can be
26 added, removed or replaced, by commands provided by btrfs device and
27 btrfs replace.
28 The profiles can be also changed, provided there's enough workspace to
30 do the conversion, using the btrfs balance command and namely the fil‐
31 ter convert.
32 Type The block group profile type is the main distinction of the in‐
34 formation stored on the block device. User data are called Data,
35 the internal data structures managed by filesystem are Metadata
36 and System.
37 Profile
39 A profile describes an allocation policy based on the redun‐
40 dancy/replication constraints in connection with the number of
41 devices. The profile applies to data and metadata block groups
42 separately. E.g. single, RAID1.
43 RAID level
45 Where applicable, the level refers to a profile that matches
46 constraints of the standard RAID levels. At the moment the sup‐
47 ported ones are: RAID0, RAID1, RAID10, RAID5 and RAID6.
48
50 Starting with a single-device filesystem
51 Assume we've created a filesystem on a block device /dev/sda with pro‐
52 file single/single (data/metadata), the device size is 50GiB and we've
53 used the whole device for the filesystem. The mount point is /mnt.
54 The amount of data stored is 16GiB, metadata have allocated 2GiB.
56 Add new device
58 We want to increase the total size of the filesystem and keep the pro‐
59 files. The size of the new device /dev/sdb is 100GiB.
60 $ btrfs device add /dev/sdb /mnt
62 The amount of free data space increases by less than 100GiB, some space
64 is allocated for metadata.
65 Convert to RAID1
67 Now we want to increase the redundancy level of both data and metadata,
68 but we'll do that in steps. Note, that the device sizes are not equal
69 and we'll use that to show the capabilities of split data/metadata and
70 independent profiles.
71 The constraint for RAID1 gives us at most 50GiB of usable space and ex‐
73 actly 2 copies will be stored on the devices.
74 First we'll convert the metadata. As the metadata occupy less than
76 50GiB and there's enough workspace for the conversion process, we can
77 do:
78 $ btrfs balance start -mconvert=raid1 /mnt
80 This operation can take a while, because all metadata have to be moved
82 and all block pointers updated. Depending on the physical locations of
83 the old and new blocks, the disk seeking is the key factor affecting
84 performance.
85 You'll note that the system block group has been also converted to
87 RAID1, this normally happens as the system block group also holds meta‐
88 data (the physical to logical mappings).
89 What changed:
91 • available data space decreased by 3GiB, usable roughly (50 - 3) +
93 (100 - 3) = 144 GiB
94 • metadata redundancy increased
96 IOW, the unequal device sizes allow for combined space for data yet im‐
98 proved redundancy for metadata. If we decide to increase redundancy of
99 data as well, we're going to lose 50GiB of the second device for obvi‐
100 ous reasons.
101 $ btrfs balance start -dconvert=raid1 /mnt
103 The balance process needs some workspace (i.e. a free device space
105 without any data or metadata block groups) so the command could fail if
106 there's too much data or the block groups occupy the whole first de‐
107 vice.
108 The device size of /dev/sdb as seen by the filesystem remains un‐
110 changed, but the logical space from 50-100GiB will be unused.
111 Remove device
113 Device removal must satisfy the profile constraints, otherwise the com‐
114 mand fails. For example:
115 $ btrfs device remove /dev/sda /mnt
117 ERROR: error removing device '/dev/sda': unable to go below two devices on raid1
118 In order to remove a device, you need to convert the profile in this
120 case:
121 $ btrfs balance start -mconvert=dup -dconvert=single /mnt
123 $ btrfs device remove /dev/sda /mnt
124
126 add [-Kf] <device> [<device>...] <path>
127 Add device(s) to the filesystem identified by path.
128 If applicable, a whole device discard (TRIM) operation is per‐
130 formed prior to adding the device. A device with existing
131 filesystem detected by blkid(8) will prevent device addition and
132 has to be forced. Alternatively the filesystem can be wiped from
133 the device using e.g. the wipefs(8) tool.
134 The operation is instant and does not affect existing data. The
136 operation merely adds the device to the filesystem structures
137 and creates some block groups headers.
138 Options
140 -K|--nodiscard
142 do not perform discard (TRIM) by default
143 -f|--force
145 force overwrite of existing filesystem on the given
146 disk(s)
147 --enqueue
149 wait if there's another exclusive operation running, oth‐
150 erwise continue
151 remove [options] <device>|<devid> [<device>|<devid>...] <path>
153 Remove device(s) from a filesystem identified by <path>
154 Device removal must satisfy the profile constraints, otherwise
156 the command fails. The filesystem must be converted to pro‐
157 file(s) that would allow the removal. This can typically happen
158 when going down from 2 devices to 1 and using the RAID1 profile.
159 See the section Typical use cases.
160 The operation can take long as it needs to move all data from
162 the device.
163 It is possible to delete the device that was used to mount the
165 filesystem. The device entry in the mount table will be replaced
166 by another device name with the lowest device id.
167 If the filesystem is mounted in degraded mode (-o degraded),
169 special term missing can be used for device. In that case, the
170 first device that is described by the filesystem metadata, but
171 not present at the mount time will be removed.
172 NOTE:
174 In most cases, there is only one missing device in degraded
175 mode, otherwise mount fails. If there are two or more devices
176 missing (e.g. possible in RAID6), you need specify missing as
177 many times as the number of missing devices to remove all of
178 them.
179 Options
181 --enqueue
183 wait if there's another exclusive operation running, oth‐
184 erwise continue
185 delete <device>|<devid> [<device>|<devid>...] <path>
187 Alias of remove kept for backward compatibility
188 replace <command> [options] <path>
190 Alias of whole command group btrfs replace for convenience. See
191 btrfs-replace(8).
192 ready <device>
194 Wait until all devices of a multiple-device filesystem are
195 scanned and registered within the kernel module. This is to pro‐
196 vide a way for automatic filesystem mounting tools to wait be‐
197 fore the mount can start. The device scan is only one of the
198 preconditions and the mount can fail for other reasons. Normal
199 users usually do not need this command and may safely ignore it.
200 scan [options] [<device> [<device>...]]
202 Scan devices for a btrfs filesystem and register them with the
203 kernel module. This allows mounting multiple-device filesystem
204 by specifying just one from the whole group.
205 If no devices are passed, all block devices that blkid reports
207 to contain btrfs are scanned.
208 The options --all-devices or -d can be used as a fallback in
210 case blkid is not available. If used, behavior is the same as
211 if no devices are passed.
212 The command can be run repeatedly. Devices that have been al‐
214 ready registered remain as such. Reloading the kernel module
215 will drop this information. There's an alternative way of mount‐
216 ing multiple-device filesystem without the need for prior scan‐
217 ning. See the mount option device.
218 Options
220 -d|--all-devices
222 Enumerate and register all devices, use as a fallback in
223 case blkid is not available.
224 -u|--forget
226 Unregister a given device or all stale devices if no path
227 is given, the device must be unmounted otherwise it's an
228 error.
229 stats [options] <path>|<device>
231 Read and print the device IO error statistics for all devices of
232 the given filesystem identified by path or for a single device.
233 The filesystem must be mounted. See section DEVICE STATS for
234 more information about the reported statistics and the meaning.
235 Options
237 -z|--reset
239 Print the stats and reset the values to zero afterwards.
240 -c|--check
242 Check if the stats are all zeros and return 0 if it is
243 so. Set bit 6 of the return code if any of the statistics
244 is no-zero. The error values is 65 if reading stats from
245 at least one device failed, otherwise it's 64.
246 -T Print stats in a tabular form, devices as rows and stats
248 as columns
249 usage [options] <path> [<path>...]::
251 Show detailed information about internal allocations on devices.
252 The level of detail can differ if the command is run under a
254 regular or the root user (due to use of restricted ioctls). The
255 first example below is for normal user (warning included) and
256 the next one with root on the same filesystem:
257 WARNING: cannot read detailed chunk info, per-device usage will not be shown, run as root
259 /dev/sdc1, ID: 1
260 Device size: 931.51GiB
261 Device slack: 0.00B
262 Unallocated: 931.51GiB
263 /dev/sdc1, ID: 1
265 Device size: 931.51GiB
266 Device slack: 0.00B
267 Data,single: 641.00GiB
268 Data,RAID0/3: 1.00GiB
269 Metadata,single: 19.00GiB
270 System,single: 32.00MiB
271 Unallocated: 271.48GiB
272 • Device size -- size of the device as seen by the filesystem
274 (may be different than actual device size)
275 • Device slack -- portion of device not used by the filesystem
277 but still available in the physical space provided by the de‐
278 vice, e.g. after a device shrink
279 • Data,single, Metadata,single, System,single -- in general,
281 list of block group type (Data, Metadata, System) and profile
282 (single, RAID1, ...) allocated on the device
283 • Data,RAID0/3 -- in particular, striped profiles
285 RAID0/RAID10/RAID5/RAID6 with the number of devices on which
286 the stripes are allocated, multiple occurrences of the same
287 profile can appear in case a new device has been added and all
288 new available stripes have been used for writes
289 • Unallocated -- remaining space that the filesystem can still
291 use for new block groups
292 Options
294 -b|--raw
296 raw numbers in bytes, without the B suffix
297 -h|--human-readable
299 print human friendly numbers, base 1024, this is the de‐
300 fault
301 -H print human friendly numbers, base 1000
303 --iec select the 1024 base for the following options, according
305 to the IEC standard
306 --si select the 1000 base for the following options, according
308 to the SI standard
309 -k|--kbytes
311 show sizes in KiB, or kB with --si
312 -m|--mbytes
314 show sizes in MiB, or MB with --si
315 -g|--gbytes
317 show sizes in GiB, or GB with --si
318 -t|--tbytes
320 show sizes in TiB, or TB with --si
321 If conflicting options are passed, the last one takes prece‐
323 dence.
324
326 The device stats keep persistent record of several error classes re‐
327 lated to doing IO. The current values are printed at mount time and up‐
328 dated during filesystem lifetime or from a scrub run.
329 $ btrfs device stats /dev/sda3
331 [/dev/sda3].write_io_errs 0
332 [/dev/sda3].read_io_errs 0
333 [/dev/sda3].flush_io_errs 0
334 [/dev/sda3].corruption_errs 0
335 [/dev/sda3].generation_errs 0
336 write_io_errs
338 Failed writes to the block devices, means that the layers be‐
339 neath the filesystem were not able to satisfy the write request.
340 read_io_errors
342 Read request analogy to write_io_errs.
343 flush_io_errs
345 Number of failed writes with the FLUSH flag set. The flushing is
346 a method of forcing a particular order between write requests
347 and is crucial for implementing crash consistency. In case of
348 btrfs, all the metadata blocks must be permanently stored on the
349 block device before the superblock is written.
350 corruption_errs
352 A block checksum mismatched or a corrupted metadata header was
353 found.
354 generation_errs
356 The block generation does not match the expected value (e.g.
357 stored in the parent node).
358 Since kernel 5.14 the device stats are also available in textual form
360 in /sys/fs/btrfs/FSID/devinfo/DEVID/error_stats.
361
363 btrfs device returns a zero exit status if it succeeds. Non zero is re‐
364 turned in case of failure.
365 If the -c option is used, btrfs device stats will add 64 to the exit
367 status if any of the error counters is non-zero.
368
370 btrfs is part of btrfs-progs. Please refer to the documentation at
371 https://btrfs.readthedocs.io.
372
374 btrfs-balance(8) btrfs-device(8), btrfs-replace(8), mkfs.btrfs(8)
3756.6.2 Nov 24, 2023 BTRFS-DEVICE(8)