wiki-grav/pages/04.other/btrfs/default.en.md

3.7 KiB

title visible
BTRFS true

[toc]

Utils

# pacman -S btrfs-progs

Fstab example

UUID=2dc70a6e-b4cf-4d94-b326-0ba9f886cf49 /mnt/tmp btrfs defaults,noatime,compress-force=zstd,space_cache=v2,subvol=@ 0 0

Options:

defaults       -> Use whatever defaults
noatime        -> Reading access to a file is not recorded
compress-force -> Send all files through the compression algorithm
space_cache    -> Increases performance by mapping unallocated blocks
subvol         -> Subvolume to mount

Filesystem usage

Show storage allocated, used and free
# btrfs fi usage (mountpoint)

allocated: space used
unallocated: actual free space
Used: amount of data stored
Free: free storage based on "Used"

Start rebalance of datachunks filled less than 70%
# btrfs balance start --b -dusage=70 -musage=70 (mountpoint)

Check status of rebalance
# btrfs balance status -v (mountpoint)

Disable CoW

Disable copy on write for folders (Only works on new files)
$ chattr +C (path)

Device errors

Error counts for a given mountpoint
# btrfs dev stat (mountpoint)

Compression

Algorithms

zlib: Slow, but strong compression, level 1-9
lzo : Fastest, weak compression
zstd: [Recommended] Medium, newer compression standard than the others, only works on newer kernels, level 1-15

Enable compression for existing files
# btrfs filesystem defragment -r -v -c(alg) (path)
It is impossible to specify the level of compression wanted.

Add compress=(alg) to /etc/fstab

To specify a level of compression (zlib and zstd) use compress=(alg):(level) in fstab.
For zstd compression it is recommended to use compress-force=zstd:(level)

Subvolumes

List
# btrfs subv list (path)

Create
# btrfs subv create (path)

Mount a subvolume
# mount -o subvol=@(subvolname) /dev/sdXX /(mountpoint)

Snapshots

TODO

RAID

An array can be mounted by specifying one of its members.
# mount /dev/sdXX /mnt

All members of an array have the same UUID, which can be mounted through fstab.

RAID 1

On filesystem creation
# mkfs.btrfs -m raid1 -d raid1 /dev/sdXX /dev/sdYY

RAID 5

On filesystem creation
It is recommended not to use raid5/6 for metadata yet
# mkfs.btrfs -m raid1 -d raid5 /dev/sdXX /dev/sdYY /dev/sdZZ

RAID 10

On filesystem creation
# mkfs.btrfs -m raid10 -d raid10 /dev/sdXX /dev/sdYY /dev/sdZZ /dev/sdQQ

Convert to single device

First, the files have to be collected on one device.
DUP on system and metadata should only be used on HDDs. Use single on SSDs
# btrfs balance start -f -sconvert=dup,devid=(id) -mconvert=dup,devid=(id) -dconvert=single,devid=(id) /(mountpoint)

Now unused devices can be removed
# btrfs device delete /dev/sdYY /(mountpoint)

Replace dying/dead device in RAID array

Show arrays that are available
btrfs fi show

From my testing the log has to be dropped before btrfs will mount the incomplete array
btrfs rescue zero-log /dev/sdXX

Mount with these options to be able to fix it
mount -o rw,degraded /(mountpoint)

The ID has to be replaced with the ID of the missing device!
btrfs replace start -B (id) /dev/sdYY /(mountpoint)

Query the status of the repace
btrfs replace status /(mountpoint)

Balance the filesystem at the end
btrfs balance /(mountpoint)

Issues

100% CPU Usage

btrfs-transaction and btrfs-cleaner will run on a single cpu core, maxing it out with 100% load.
TODO: Check what enabled quotas in the first place. A likely candidate is snapper
The issue is apparently caused by using quotas in btrfs.
Check if quotas are enabled:
# btrfs qgroup show (path)
Disable quotas:
# btrfs quota disable (path)