Disable Copy-On-Write on BTRFS
The issue of COW (copy on write), is fragmentation, because it always write new device block. This is good for SSD, but not good on traditional devices. Even on SSD, if the block size is big, the data to be write would be much larger than actual updated data size. Because of this issue, recommented disable Copy-On-Write on database and VM filesystems.
Disable it by mounting with nodatacow.
Following facts to be considered
- This implies nodatasum as well
- COW may still happen if a snapshot is taken
- COW will still be maintained for existing files
- COW status can be modified only for empty or newly created files.
For an empty file, add the NOCOW file attribute (use chattr utility with +C)
touch file1 chattr +C file1
For a directory with NOCOW attribute set, new files in it will inherit this attribute.
chattr +C directory1
For old files, copy the original data into the pre-created file, delete original and rename back.
touch vm-image.raw chattr +C vm-image.raw fallocate -l10g vm-image.raw
Subvolume can not be set nocow separately. This is official answer.
But the files created inherit the attributes from directory, if separately mount subvolume on the directory which has nocow attribute, then the newly created files will inherit nocow attribute as well, regardless of the original volume.
mkdir /var/lib/nocow chattr +C /var/lib/nocow
mount -o autodefrag,compress=lzo,noatime,space_cache /dev/mapper/zpool1 /mnt/zpool1 btrfs subvolume create /mnt/zpool1/nocow
/dev/mapper/zpool1 /var/lib/nocow btrfs rw,noatime,compress=lzo,space_cache,autodefrag,subvol=nocow 0 0
No checksum, no integrity.
Nodatacow bypasses the very mechanisms that are meant to provide consistency in the filesystem, because the CoW operations are achieved by constructing a completely new metadata tree containing both changes (references to the data, and the csum metadata), and then atomically changing the superblock to point to the new tree.
With nodatacow, writing data and checksum on the physical medium, cause two writes separately. This could cause the data and the checksum mismatch due to I/O error, file corruption could happen.