Provided by: btrfs-progs_6.6.3-1.1build2_amd64 bug

NAME

       btrfs-scrub - scrub btrfs filesystem, verify block checksums

SYNOPSIS

       btrfs scrub <subcommand> <args>

DESCRIPTION

       Scrub  is  a  pass over all filesystem data and metadata and verifying the checksums. If a
       valid copy is available  (replicated  block  group  profiles)  then  the  damaged  one  is
       repaired. All copies of the replicated profiles are validated.

       NOTE:
          Scrub  is  not  a  filesystem  checker (fsck) and does not verify nor repair structural
          damage in the filesystem. It really only checks checksums of data and tree  blocks,  it
          doesn't  ensure  the  content  of  tree  blocks  is  valid and consistent. There's some
          validation performed when metadata blocks are read from disk (Tree  checker)  but  it's
          not extensive and cannot substitute full btrfs-check(8) run.

       The  user is supposed to run it manually or via a periodic system service. The recommended
       period is a month but it could be less. The  estimated  device  bandwidth  utilization  is
       about 80% on an idle filesystem.

       The   scrubbing   status   is   recorded   in   /var/lib/btrfs/  in  textual  files  named
       scrub.status.UUID for a filesystem identified  by  the  given  UUID.  (Progress  state  is
       communicated  through a named pipe in file scrub.progress.UUID in the same directory.) The
       status file is updated every 5 seconds. A resumed scrub will continue from the last  saved
       position.

       Scrub  can  be  started only on a mounted filesystem, though it's possible to scrub only a
       selected device. See btrfs scrub start for more.

   Bandwidth and IO limiting
       NOTE:
          The ionice(1) may not be generally supported by all IO schedulers and  the  options  to
          btrfs scrub start may not work as expected.

       In  the  past  when the CFQ IO scheduler was generally used the ionice(1) syscalls set the
       priority to idle so the IO would not interfere with regular IO. Since the kernel  5.0  the
       CFQ is not available.

       The  IO  scheduler  known  to support that is BFQ, but first read the documentation before
       using it!

       For other commonly used schedulers like mq-deadline it's recommended  to  use  cgroup2  IO
       controller    which    could    be    managed    by    e.g.    systemd    (documented   in
       systemd.resource-control). However,  starting  scrub  like  that  is  not  yet  completely
       straightforward.  The  IO  controller  must know the physical device of the filesystem and
       create a slice so all processes started from that belong to the same accounting group.

          $ systemd-run -p "IOBandwidthReadMax=/dev/sdx 10M" btrfs scrub start -B /

       Since linux 5.14 it's possible to set the per-device bandwidth limits in a  BTRFS-specific
       way  using  files  /sys/fs/btrfs/FSID/devinfo/DEVID/scrub_speed_max.   This setting is not
       persistent, lasts until  the  filesystem  is  unmounted.   Currently  set  limits  can  be
       displayed by command btrfs scrub limit.

          $ echo 100m > /sys/fs/btrfs/9b5fd16e-1b64-4f9b-904a-74e74c0bbadc/devinfo/1/scrub_speed_max
          $ btrfs scrub limit /
          UUID: 9b5fd16e-1b64-4f9b-904a-74e74c0bbadc
          Id      Limit      Path
          --  ---------  --------
           1  100.00MiB  /dev/sdx

SUBCOMMAND

       cancel <path>|<device>
              If a scrub is running on the filesystem identified by path or device, cancel it.

              If  a  device  is  specified, the corresponding filesystem is found and btrfs scrub
              cancel behaves as if it was called on that filesystem.  The progress  is  saved  in
              the status file so btrfs scrub resume can continue from the last position.

       limit [options] <path>
              Show or set scrub limits on devices of the given filesystem.

              Options

              -d|--devid DEVID
                     select the device by DEVID to apply the limit

              -l|--limit SIZE
                     set  the limit of the device to SIZE (size units with suffix), or 0 to reset
                     to unlimited

              -a|--all
                     apply the limit to all devices

              --raw  print all numbers raw values in bytes without the B suffix

              --human-readable
                     print human friendly numbers, base 1024, this is the default

              --iec  select the 1024 base  for  the  following  options,  according  to  the  IEC
                     standard

              --si   select the 1000 base for the following options, according to the SI standard

              --kbytes
                     show sizes in KiB, or kB with --si

              --mbytes
                     show sizes in MiB, or MB with --si

              --gbytes
                     show sizes in GiB, or GB with --si

              --tbytes
                     show sizes in TiB, or TB with --si

       resume [-BdqrR] <path>|<device>
              Resume  a cancelled or interrupted scrub on the filesystem identified by path or on
              a given device. The starting point is read from the status file if it exists.

              This does not start a new scrub if the last scrub finished successfully.

              Options

              see scrub start.

       start [-BdrRf] <path>|<device>
              Start a scrub on all devices of the mounted filesystem identified by path or  on  a
              single  device. If a scrub is already running, the new one will not start. A device
              of an unmounted filesystem cannot be scrubbed this way.

              Without options, scrub is started as a background process. The automatic repairs of
              damaged  copies  are performed by default for block group profiles with redundancy.
              No-repair can be enabled by option -r.

              Options

              -B     do not background and print scrub statistics when finished

              -d     print separate statistics for each device of the filesystem (-B only) at the
                     end

              -r     run  in  read-only mode, do not attempt to correct anything, can be run on a
                     read-only filesystem

              -R     raw print mode, print full data instead of summary

              -f     force starting new scrub even if a scrub is already running, this can useful
                     when scrub status file is damaged and reports a running scrub although it is
                     not, but should not normally be necessary

              Deprecated options

              -c <ioprio_class>
                     set IO priority class (see  ionice(1)  manual  page)  if  the  IO  scheduler
                     configured  for the device supports ionice. This is only supported by BFQ or
                     Kyber but is not supported by mq-deadline. Please read the section about  IO
                     limiting.

              -n <ioprio_classdata>
                     set IO priority classdata (see ionice(1) manpage)

              -q     (deprecated) alias for global -q option

       status [options] <path>|<device>
              Show  status  of  a  running scrub for the filesystem identified by path or for the
              specified device.

              If no scrub is running, show statistics of the last finished or cancelled scrub for
              that filesystem or device.

              Options

              -d     print separate statistics for each device of the filesystem

              -R     print  all  raw  statistics without postprocessing as returned by the status
                     ioctl

              --raw  print all numbers raw values in bytes without the B suffix

              --human-readable
                     print human friendly numbers, base 1024, this is the default

              --iec  select the 1024 base  for  the  following  options,  according  to  the  IEC
                     standard

              --si   select the 1000 base for the following options, according to the SI standard

              --kbytes
                     show sizes in KiB, or kB with --si

              --mbytes
                     show sizes in MiB, or MB with --si

              --gbytes
                     show sizes in GiB, or GB with --si

              --tbytes
                     show sizes in TiB, or TB with --si

              A status on a filesystem without any error looks like the following:

                 # btrfs scrub start /
                 # btrfs scrub status /
                 UUID:             76fac721-2294-4f89-a1af-620cde7a1980
                 Scrub started:    Wed Apr 10 12:34:56 2023
                 Status:           running
                 Duration:         0:00:05
                 Time left:        0:00:05
                 ETA:              Wed Apr 10 12:35:01 2023
                 Total to scrub:   28.32GiB
                 Bytes scrubbed:   13.76GiB  (48.59%)
                 Rate:             2.75GiB/s
                 Error summary:    no errors found

              With some errors found:

                 Error summary:    csum=72
                   Corrected:      2
                   Uncorrectable:  72
                   Unverified:     0

              • Corrected -- number of bad blocks that were repaired from another copy

              • Uncorrectable  --  errors  detected  at read time but not possible to repair from
                other copy

              • Unverified -- transient errors, first read failed but a retry succeeded,  may  be
                affected by lower layers that group or split IO requests

              • Error summary -- followed by a more detailed list of errors found

                • csum -- checksum mismatch

                • super  --  super  block errors, unless the error is fixed immediately, the next
                  commit will overwrite superblock

                • verify -- metadata block header errors

                • read -- blocks can't be read due to IO errors

              It's     possible     to     set     a     per-device      limit      via      file
              sysfs/fs/btrfs/FSID/devinfo/scrub_speed_max.  In  that case the limit is printed on
              the Rate: line if option  -d  is  specified,  or  without  it  on  a  single-device
              filesystem.  Read more about tat in section about scrub IO limiting.

                 Rate:             989.0MiB/s (limit 1.0G/s)

              On  a  multi-device  filesystem  with  at  least one device limit the overall stats
              cannot print the limit without -d so there's a not that some limits are set:

                 Rate:             36.37MiB/s (some device limits set)

EXIT STATUS

       btrfs scrub returns a zero exit status if it succeeds. Non zero is  returned  in  case  of
       failure:

       1      scrub couldn't be performed

       2      there is nothing to resume

       3      scrub found uncorrectable errors

AVAILABILITY

       btrfs    is    part    of   btrfs-progs.    Please   refer   to   the   documentation   at
       https://btrfs.readthedocs.io.

SEE ALSO

       mkfs.btrfs(8)