6 btrfs-balance - balance block groups on a btrfs filesystem
10 *btrfs balance* <subcommand> <args>
14 The primary purpose of the balance feature is to spread block groups across
15 all devices so they match constraints defined by the respective profiles. See
16 `mkfs.btrfs`(8) section 'PROFILES' for more details.
17 The scope of the balancing process can be further tuned by use of filters that
18 can select the block groups to process. Balance works only on a mounted
21 The balance operation is cancellable by the user. The on-disk state of the
22 filesystem is always consistent so an unexpected interruption (eg. system crash,
23 reboot) does not corrupt the filesystem. The progress of the balance operation
24 is temporarily stored and will be resumed upon mount, unless the mount option
25 'skip_balance' is specified.
27 WARNING: running balance without filters will take a lot of time as it basically
28 rewrites the entire filesystem and needs to update all block pointers.
30 The filters can be used to perform following actions:
32 - convert block group profiles (filter 'convert')
33 - make block group usage more compact (filter 'usage')
34 - perform actions only on a given device (filters 'devid', 'drange')
36 The filters can be applied to a combination of block group types (data,
37 metadata, system). Note that changing 'system' needs the force option.
39 NOTE: the balance operation needs enough work space, ie. space that is
40 completely unused in the filesystem, otherwise this may lead to ENOSPC reports.
41 See the section 'ENOSPC' for more details.
46 NOTE: The balance subcommand also exists under the *btrfs filesystem*
47 namespace. This still works for backward compatibility but is deprecated and
48 should not be used anymore.
50 NOTE: A short syntax *btrfs balance <path>* works due to backward compatibility
51 but is deprecated and should not be used anymore. Use *btrfs balance start*
54 PERFORMANCE IMPLICATIONS
55 ------------------------
57 Balance operation is intense namely in the IO respect, but can be also CPU
58 intense. It affects other actions on the filesystem. There are typically lots
59 of data being copied from one location to another, and lots of metadata get
62 Depending on the actual block group layout, it can be also seek-heavy. The
63 performance on rotational devices is noticeably worse than on SSDs or fast
69 cancel running or paused balance, the command will block and wait until the
70 actually processed blockgroup is finished
73 pause running balance operation, this will store the state of the balance
74 progress and used filters to the filesystem
77 resume interrupted balance, the balance status must be stored on the filesystem
78 from previous run, eg. after it was forcibly interrupted and mounted again with
81 *start* [options] <path>::
82 start the balance operation according to the specified filters, no filters
83 will rewrite the entire filesystem. The process runs in the foreground.
85 NOTE: the balance command without filters will basically rewrite everything
86 in the filesystem. The run time is potentially very long, depending on the
87 filesystem size. To prevent starting a full balance by accident, the user is
88 warned and has a few seconds to cancel the operation before it starts. The
89 warning and delay can be skipped with '--full-balance' option.
91 Please note that the filters must be written together with the '-d', '-m' and
92 '-s' options, because they're optional and bare '-d' etc alwo work and mean no
98 act on data block groups, see `FILTERS` section for details about 'filters'
100 act on metadata chunks, see `FILTERS` section for details about 'filters'
102 act on system chunks (requires '-f'), see `FILTERS` section for details about 'filters'.
104 be verbose and print balance filter arguments
106 force reducing of metadata integrity, eg. when going from 'raid1' to 'single'
108 *status* [-v] <path>::
109 Show status of running or paused balance.
111 If '-v' option is given, output will be verbose.
115 From kernel 3.3 onwards, btrfs balance can limit its action to a subset of the
116 whole filesystem, and can be used to change the replication configuration (e.g.
117 moving data from single to RAID1). This functionality is accessed through the
118 '-d', '-m' or '-s' options to btrfs balance start, which filter on data,
119 metadata and system blocks respectively.
121 A filter has the following structure: 'type'[='params'][,'type'=...]
123 The available types are:
125 *profiles=<profiles>*::
126 Balances only block groups with the given profiles. Parameters
127 are a list of profile names separated by "'|'" (pipe).
131 Balances only block groups with usage under the given percentage. The
132 value of 0 is allowed and will clean up completely unused block groups, this
133 should not require any new work space allocated. You may want to use 'usage=0'
134 in case balance is returning ENOSPC and your filesystem is not too full.
136 The argument may be a single value or a range. The single value 'N' means 'at
137 most N percent used', equivalent to '..N' range syntax. Kernels prior to 4.4
138 accept only the single value format.
139 The minimum range boundary is inclusive, maximum is exclusive.
142 Balances only block groups which have at least one chunk on the given
143 device. To list devices with ids use *btrfs fi show*.
146 Balance only block groups which overlap with the given byte range on any
147 device. Use in conjunction with 'devid' to filter on a specific device. The
148 parameter is a range specified as 'start..end'.
151 Balance only block groups which overlap with the given byte range in the
152 filesystem's internal virtual address space. This is the address space that
153 most reports from btrfs in the kernel log use. The parameter is a range
154 specified as 'start..end'.
156 *convert=<profile>*::
157 Convert each selected block group to the given profile name identified by
160 NOTE: starting with kernel 4.5, the 'data' chunks can be converted to/from the
161 'DUP' profile on a single device.
163 NOTE: starting with kernel 4.6, all profiles can be converted to/from 'DUP' on
164 multi-device filesystems.
168 Process only given number of chunks, after all filters are applied. This can be
169 used to specifically target a chunk in connection with other filters ('drange',
170 'vrange') or just simply limit the amount of work done by a single balance run.
172 The argument may be a single value or a range. The single value 'N' means 'at
173 most N chunks', equivalent to '..N' range syntax. Kernels prior to 4.4 accept
174 only the single value format. The range minimum and maximum are inclusive.
177 Balance only block groups which have the given number of stripes. The parameter
178 is a range specified as 'start..end'. Makes sense for block group profiles that
179 utilize striping, ie. RAID0/10/5/6. The range minimum and maximum are
183 Takes no parameters. Only has meaning when converting between profiles.
184 When doing convert from one profile to another and soft mode is on,
185 chunks that already have the target profile are left untouched.
186 This is useful e.g. when half of the filesystem was converted earlier but got
189 The soft mode switch is (like every other filter) per-type.
190 For example, this means that we can convert metadata chunks the "hard" way
191 while converting data chunks selectively with soft switch.
193 Profile names, used in 'profiles' and 'convert' are one of: 'raid0', 'raid1',
194 'raid10', 'raid5', 'raid6', 'dup', 'single'. The mixed data/metadata profiles
195 can be converted in the same way, but it's conversion between mixed and non-mixed
196 is not implemented. For the constraints of the profiles please refer to `mkfs.btrfs`(8),
202 The way balance operates, it usually needs to temporarily create a new block
203 group and move the old data there. For that it needs work space, otherwise
204 it fails for ENOSPC reasons.
205 This is not the same ENOSPC as if the free space is exhausted. This refers to
206 the space on the level of block groups.
208 The free work space can be calculated from the output of the *btrfs filesystem show*
211 ------------------------------
212 Label: 'BTRFS' uuid: 8a9d72cd-ead3-469d-b371-9c7203276265
213 Total devices 2 FS bytes used 77.03GiB
214 devid 1 size 53.90GiB used 51.90GiB path /dev/sdc2
215 devid 2 size 53.90GiB used 51.90GiB path /dev/sde1
216 ------------------------------
218 'size' - 'used' = 'free work space' +
219 '53.90GiB' - '51.90GiB' = '2.00GiB'
221 An example of a filter that does not require workspace is 'usage=0'. This will
222 scan through all unused block groups of a given type and will reclaim the
223 space. After that it might be possible to run other filters.
225 **CONVERSIONS ON MULTIPLE DEVICES**
227 Conversion to profiles based on striping (RAID0, RAID5/6) require the work
228 space on each device. An interrupted balance may leave partially filled block
229 groups that might consume the work space.
234 A more comprehensive example when going from one to multiple devices, and back,
235 can be found in section 'TYPICAL USECASES' of `btrfs-device`(8).
237 MAKING BLOCK GROUP LAYOUT MORE COMPACT
238 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
240 The layout of block groups is not normally visible, most tools report only
241 summarized numbers of free or used space, but there are still some hints
244 Let's use the following real life example and start with the output:
248 Data, single: total=75.81GiB, used=64.44GiB
249 System, RAID1: total=32.00MiB, used=20.00KiB
250 Metadata, RAID1: total=15.87GiB, used=8.84GiB
251 GlobalReserve, single: total=512.00MiB, used=0.00B
254 Roughly calculating for data, '75G - 64G = 11G', the used/total ratio is
255 about '85%'. How can we can interpret that:
257 * chunks are filled by 85% on average, ie. the 'usage' filter with anything
258 smaller than 85 will likely not affect anything
259 * in a more realistic scenario, the space is distributed unevenly, we can
260 assume there are completely used chunks and the remaining are partially filled
262 Compacting the layout could be used on both. In the former case it would spread
263 data of a given chunk to the others and removing it. Here we can estimate that
264 roughly 850 MiB of data have to be moved (85% of a 1 GiB chunk).
266 In the latter case, targeting the partially used chunks will have to move less
267 data and thus will be faster. A typical filter command would look like:
270 # btrfs balance start -dusage=50 /path
271 Done, had to relocate 2 out of 97 chunks
274 Data, single: total=74.03GiB, used=64.43GiB
275 System, RAID1: total=32.00MiB, used=20.00KiB
276 Metadata, RAID1: total=15.87GiB, used=8.84GiB
277 GlobalReserve, single: total=512.00MiB, used=0.00B
280 As you can see, the 'total' amount of data is decreased by just 1 GiB, which is
281 an expected result. Let's see what will happen when we increase the estimated
285 # btrfs balance start -dusage=85 /path
286 Done, had to relocate 13 out of 95 chunks
289 Data, single: total=68.03GiB, used=64.43GiB
290 System, RAID1: total=32.00MiB, used=20.00KiB
291 Metadata, RAID1: total=15.87GiB, used=8.85GiB
292 GlobalReserve, single: total=512.00MiB, used=0.00B
295 Now the used/total ratio is about 94% and we moved about '74G - 68G = 6G' of
296 data to the remaining blockgroups, ie. the 6GiB are now free of filesystem
297 structures, and can be reused for new data or metadata block groups.
299 We can do a similar exercise with the metadata block groups, but this should
300 not be typically necessary, unless the used/total ration is really off. Here
301 the ratio is roughly 50% but the difference as an absolute number is "a few
302 gigabytes", which can be considered normal for a workload with snapshots or
303 reflinks updated frequently.
306 # btrfs balance start -musage=50 /path
307 Done, had to relocate 4 out of 89 chunks
310 Data, single: total=68.03GiB, used=64.43GiB
311 System, RAID1: total=32.00MiB, used=20.00KiB
312 Metadata, RAID1: total=14.87GiB, used=8.85GiB
313 GlobalReserve, single: total=512.00MiB, used=0.00B
316 Just 1 GiB decrease, which possibly means there are block groups with good
317 utilization. Making the metadata layout more compact would in turn require
318 updating more metadata structures, ie. lots of IO. As running out of metadata
319 space is a more severe problem, it's not necessary to keep the utilization
320 ratio too high. For the purpose of this example, let's see the effects of
324 # btrfs balance start -musage=70 /path
325 Done, had to relocate 13 out of 88 chunks
328 Data, single: total=68.03GiB, used=64.43GiB
329 System, RAID1: total=32.00MiB, used=20.00KiB
330 Metadata, RAID1: total=11.97GiB, used=8.83GiB
331 GlobalReserve, single: total=512.00MiB, used=0.00B
334 GETTING RID OF COMPLETELY UNUSED BLOCK GROUPS
335 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
337 Normally the balance operation needs a work space, to temporarily move the
338 data before the old block groups gets removed. If there's no work space, it
339 ends with 'no space left'.
341 There's a special case when the block groups are completely unused, possibly
342 left after removing lots of files or deleting snapshots. Removing empty block
343 groups is automatic since 3.18. The same can be achieved manually with a
344 notable exception that this operation does not require the work space. Thus it
345 can be used to reclaim unused block groups to make it available.
348 # btrfs balance start -dusage=0 /path
351 This should lead to decrease in the 'total' numbers in the *btrfs fi df* output.
355 *btrfs balance* returns a zero exit status if it succeeds. Non zero is
356 returned in case of failure.
360 *btrfs* is part of btrfs-progs.
361 Please refer to the btrfs wiki http://btrfs.wiki.kernel.org for
projects / platform / upstream / btrfs-progs.git / blob