Packing zfs snapshots with tar on running container

[dominikp]

I just want to clear out some things.

I have created some dataset on host.

rpool/data/101-jenkinsfolder

This dataset is mounted in container

mp0: /rpool/data/101-jenkinsfolder,mp=/mnt/101-jenkinsfolder

Because proxmox wont backup it, i need to do it manually. I would want to do it like this:

zfs snapshot rpool/data/101-jenkinsfolder@snapshot_name
cd /rpool/data/101-jenkinsfolder/.zfs/snapshot
tar -cvzf /some_backup_dir/snapshot_name.tar.gz snapshot_name/

QUESTIONS
1. I assume that content of snapshot is frozen so there should be no problem to perform this process on running machine? Because thats why i want to make snapshot before packing.
2. Snapshots are incremental so i dont have to worry about their number? I dont have to delete old ones because then i will destroy the "base" data for newer snapshots?

 

[fiona]

QUESTIONS
1. I assume that content of snapshot is frozen so there should be no problem to perform this process on running machine? Because thats why i
want to make snapshot before packing.

Yes. If nothing weird happens, the filesystems in .zfs/snapshot are read-only.

2. Snapshots are incremental so i dont have to worry about their number? I dont have to delete old ones because then i will destroy the "base" data for newer snapshots?

You can delete older snapshots, while keeping newer ones (and vice versa). ZFS handles resolving the differences internally. If a dataset really depends on another one and you want to destroy that one, ZFS will tell you and refuse to delete it (except if you pass the recursive switch -r).

 

[dominikp]

Thanks for reply!

You can delete older snapshots, while keeping newer ones (and vice versa). ZFS handles resolving the differences internally. If a dataset really depends on another one and you want to destroy that one, ZFS will tell you and refuse to delete it (except if you pass the recursive switch -r).

Ok, but it doesnt make much sense right? I mean, deleting old snapshots wont give you additional space?

 

[fiona]

You should always get at least a little bit of space back and if some of the changes are over-written by others you can gain more space. For example:

root@pve /myzpool/myfs # zfs list -o space -r -t all myzpool/myfs
NAME                         AVAIL   USED  USEDSNAP  USEDDS  USEDREFRESERV  USEDCHILD
myzpool/myfs                 21.4G  39.2M        0B   39.2M             0B         0B
myzpool/myfs@snap1               -     0B         -       -              -          -

root@pve /myzpool/myfs # dd if=/dev/urandom of=file count=10000 bs=4K    
10000+0 records in
10000+0 records out
40960000 bytes (41 MB, 39 MiB) copied, 0.349116 s, 117 MB/s

root@pve /myzpool/myfs # zfs list -o space -r -t all myzpool/myfs
NAME                         AVAIL   USED  USEDSNAP  USEDDS  USEDREFRESERV  USEDCHILD
myzpool/myfs                 21.4G  78.4M     39.2M   39.2M             0B         0B
myzpool/myfs@snap1               -  39.2M         -       -              -          -

root@pve /myzpool/myfs # zfs snapshot myzpool/myfs@snap2         

root@pve /myzpool/myfs # zfs list -o space -r -t all myzpool/myfs
NAME                         AVAIL   USED  USEDSNAP  USEDDS  USEDREFRESERV  USEDCHILD
myzpool/myfs                 21.4G  78.4M     39.2M   39.2M             0B         0B
myzpool/myfs@snap1               -  39.2M         -       -              -          -
myzpool/myfs@snap2               -     0B         -       -              -          -

root@pve /myzpool/myfs # dd if=/dev/urandom of=file count=10000 bs=4K     
10000+0 records in
10000+0 records out
40960000 bytes (41 MB, 39 MiB) copied, 0.24106 s, 170 MB/s

root@pve /myzpool/myfs # zfs list -o space -r -t all myzpool/myfs
NAME                         AVAIL   USED  USEDSNAP  USEDDS  USEDREFRESERV  USEDCHILD
myzpool/myfs                 21.4G   118M     78.3M   39.2M             0B         0B
myzpool/myfs@snap1               -  39.2M         -       -              -          -
myzpool/myfs@snap2               -  39.2M         -       -              -          -

root@pve /myzpool/myfs # zfs destroy myzpool/myfs@snap1 

root@pve /myzpool/myfs # zfs list -o space -r -t all myzpool/myfs
NAME                         AVAIL   USED  USEDSNAP  USEDDS  USEDREFRESERV  USEDCHILD
myzpool/myfs                 21.4G  78.4M     39.2M   39.2M             0B         0B
myzpool/myfs@snap2               -  39.2M         -       -              -          -

The USED column for a snapshot shows how much space can be gained by deleting that snapshot. The USEDSNAP column for the dataset shows how much space is used by all snapshots together.