I have a 16M file.
I take a snapshot of the ZFS filesystem which contains it.
If I overwrite the file with the same data, will ZFS need to store two copies of all of the blocks of the file?
Add a comment
|
2 Answers
Yes, ZFS will allocate extra space for the same file if a snapshot is taken from the fileset.
First let's create an empty fileset and prepare a file to copy from. For simplicity reasons, compression was turned off and the pool was created on a single disk without any raidz or mirror.
[root@localhost ~]# dd if=/dev/urandom of=/tmp/testfile bs=16M count=1
1+0 records in
1+0 records out
16777216 bytes (17 MB) copied, 0.113345 s, 148 MB/s
[root@localhost ~]# zpool create tank sdd
[root@localhost ~]# zfs create tank/test
The space used on disk can be seen with zpool list.
[root@localhost]# zpool list tank
NAME SIZE ALLOC FREE EXPANDSZ FRAG CAP DEDUP HEALTH ALTROOT
tank 9.94G 182K 9.94G - 0% 0% 1.00x ONLINE -
Now copy the file over to the ZFS fileset, create a snapshot and look at the space used.
[root@localhost ~]# /bin/cp /tmp/testfile /tank/test/
[root@localhost ~]# zfs list -t all -r tank/test
NAME USED AVAIL REFER MOUNTPOINT
tank/test 16.0M 9.61G 16.0M /tank/test
[root@localhost ~]# zfs snapshot tank/test@1
[root@localhost ~]# zfs list -t all -r tank/test
NAME USED AVAIL REFER MOUNTPOINT
tank/test 16.0M 9.61G 16.0M /tank/test
tank/test@1 0B - 16.0M -
Okay, copy the same file again to the same location in the ZFS fileset and look at the space used again.
[root@localhost ~]# /bin/cp -f /tmp/testfile /tank/test/
[root@localhost ~]# zfs list -t all -r tank/test
NAME USED AVAIL REFER MOUNTPOINT
tank/test 32.0M 9.60G 16.0M /tank/test
tank/test@1 16.0M - 16.0M -
Also the disk space used in the pool is growing to 32MB.
[root@localhost tank]# zpool list tank
NAME SIZE ALLOC FREE EXPANDSZ FRAG CAP DEDUP HEALTH ALTROOT
tank 9.94G 32.2M 9.91G - 0% 0% 1.00x ONLINE -
As you can see, the tank/test fileset now accounts for 32MB in total, splitting up in 16MB in the tank/test fileset and 16MB in the tank/test@1 snapshot. Also the output of zpool list shows an allocation of 32MB on disk.
If you repeat copy and take snapshot the tank/test total USED would grow further.
Update
Thanks @Andrew Henle for asking. Had to update my answer above and will continue the deduplication stuff below.
Let's do this again with a fileset that has deduplication enabled.
[root@localhost]# zfs destroy -r tank/test
[root@localhost]# zfs create tank/test-dedup
[root@localhost]# zfs set dedup=on tank/test-dedup
I will skip the single steps here and will just add the output of the overview of the used space.
[root@localhost ~]# zfs list -t all -r tank/test-dedup
NAME USED AVAIL REFER MOUNTPOINT
tank 32.3M 9.61G 24K /tank
tank/test-dedup 32.1M 9.61G 16.0M /tank/test-dedup
tank/test-dedup@1 16.0M - 16.0M -
[root@localhost ~]# zpool list tank
NAME SIZE ALLOC FREE EXPANDSZ FRAG CAP DEDUP HEALTH ALTROOT
tank 9.94G 16.3M 9.92G - 0% 0% 2.00x ONLINE -
With deduplication enabled, the snapshot is still reported to use the extra size, but as you can see with zpool list the deduplication is saving the space on disk and gives a deduplication ratio of 2.
IIRC in general it is not recommended to use deduplication due to heavy memory usage and the resulting performance impact. I think there is work going on to improve the deduplication feature in ZFS.
I know this is old, but (for some versions of OpenZFS, at least) you can avoid this extra allocation of identical blocks using nop-write
ZFS supports end-to-end checksumming of every data block. When a cryptographically secure checksum is being used (and compression is enabled) OpenZFS will compare the checksums of incoming writes to checksum of the existing on-disk data and avoid issuing any write i/o for data that has not changed. This can help performance and snapshot space usage in situations were the same files are regularly overwritten with almost-identical data (e.g. regular full-backups of large random-access files).
