Zfs Snapshot fails - out of space

[yena]

Hello,
on my server i can't do new snapshot:

-----------------------------------------------------------------------
root@nodo5:~# zfs snapshot rpool/KVM/vm-105-disk-1@snap1
cannot create snapshot 'rpool/KVM/vm-105-disk-1@snap1': out of space
root@nodo5:~#
root@nodo5:~# zfs list -t all
NAME USED AVAIL REFER MOUNTPOINT
rpool 1.63T 398G 73K /rpool
rpool/BACKUP 57.5K 398G 57.5K /rpool/BACKUP
rpool/KVM 1.62T 398G 57.5K /rpool/KVM
rpool/KVM/vm-105-disk-1 1.62T 1.55T 470G -
rpool/KVM/vm-105-disk-1@__base__ 4.01G - 7.86G -
rpool/LXC 57.5K 398G 57.5K /rpool/LXC
rpool/ROOT 2.51G 398G 96K /rpool/ROOT
rpool/ROOT/pve-1 2.51G 398G 2.51G /
rpool/data 96K 398G 96K /rpool/data
rpool/swap 8.50G 406G 1.35G -

-----------------------------------------------------------------------

But seem i have 1.55T ..
Any idea ?

Thanks!

 

[Klaus Steinberger]

your pool is full :

rpool 1.63T 398G 73K /rpool
rpool/KVM 1.62T 398G 57.5K /rpool/KVM

one hint: never fill up a COW Filesystem like ZFS, 100% full is really bad, even no deletion is then possible, for good performance better keep it under 70-80 %

 

[yena]

your pool is full :

rpool 1.63T 398G 73K /rpool
rpool/KVM 1.62T 398G 57.5K /rpool/KVM

one hint: never fill up a COW Filesystem like ZFS, 100% full is really bad, even no deletion is then possible, for good performance better keep it under 70-80 %

I see 398G free, where do you see is full ?
Thanks

 

[LnxBil]

Please also post (inside CODE tags for better readability)

zpool list

 

[yena]

Please also post (inside CODE tags for better readability)

zpool list

root@nodo5:~# zpool list
NAME SIZE ALLOC FREE EXPANDSZ FRAG CAP DEDUP HEALTH ALTROOT
rpool 2.09T 480G 1.62T - 34% 22% 1.00x ONLINE -


FREE: 1.62T

 

[LnxBil]

Thank you for NOT using CODE tags - just format it in a non-proportional font does not work well with spaces as you demonstrated.

Now, please also post

zfs get all rpool/KVM/vm-105-disk-1

 

[yena]

Thank you for NOT using CODE tags - just format it in a non-proportional font does not work well with spaces as you demonstrated.

Now, please also post

zfs get all rpool/KVM/vm-105-disk-1

root@nodo5:~# zfs get all rpool/KVM/vm-105-disk-1
NAME PROPERTY VALUE SOURCE
rpool/KVM/vm-105-disk-1 type volume -
rpool/KVM/vm-105-disk-1 creation Thu Aug 3 15:16 2017 -
rpool/KVM/vm-105-disk-1 used 1.62T -
rpool/KVM/vm-105-disk-1 available 1.55T -
rpool/KVM/vm-105-disk-1 referenced 470G -
rpool/KVM/vm-105-disk-1 compressratio 1.41x -
rpool/KVM/vm-105-disk-1 reservation none default
rpool/KVM/vm-105-disk-1 volsize 1.56T local
rpool/KVM/vm-105-disk-1 volblocksize 8K default
rpool/KVM/vm-105-disk-1 checksum on default
rpool/KVM/vm-105-disk-1 compression on inherited from rpool
rpool/KVM/vm-105-disk-1 readonly off default
rpool/KVM/vm-105-disk-1 createtxg 887 -
rpool/KVM/vm-105-disk-1 copies 1 default
rpool/KVM/vm-105-disk-1 refreservation 1.61T local
rpool/KVM/vm-105-disk-1 guid 4327597119344679116 -
rpool/KVM/vm-105-disk-1 primarycache all default
rpool/KVM/vm-105-disk-1 secondarycache all default
rpool/KVM/vm-105-disk-1 usedbysnapshots 4.01G -
rpool/KVM/vm-105-disk-1 usedbydataset 470G -
rpool/KVM/vm-105-disk-1 usedbychildren 0B -
rpool/KVM/vm-105-disk-1 usedbyrefreservation 1.16T -
rpool/KVM/vm-105-disk-1 logbias latency default
rpool/KVM/vm-105-disk-1 dedup off default
rpool/KVM/vm-105-disk-1 mlslabel none default
rpool/KVM/vm-105-disk-1 sync standard inherited from rpool
rpool/KVM/vm-105-disk-1 refcompressratio 1.41x -
rpool/KVM/vm-105-disk-1 written 466G -
rpool/KVM/vm-105-disk-1 logicalused 658G -
rpool/KVM/vm-105-disk-1 logicalreferenced 650G -
rpool/KVM/vm-105-disk-1 volmode default default
rpool/KVM/vm-105-disk-1 snapshot_limit none default
rpool/KVM/vm-105-disk-1 snapshot_count none default
rpool/KVM/vm-105-disk-1 snapdev hidden default
rpool/KVM/vm-105-disk-1 context none default
rpool/KVM/vm-105-disk-1 fscontext none default
rpool/KVM/vm-105-disk-1 defcontext none default
rpool/KVM/vm-105-disk-1 rootcontext none default
rpool/KVM/vm-105-disk-1 redundant_metadata all default

Thanks!

 

[LnxBil]

You set refreservation, unset and it'll most probably work.

Please read this article:
http://www.mceith.com/blog/?p=153

 

[yena]

You set refreservation, unset and it'll most probably work.

Please read this article:
http://www.mceith.com/blog/?p=153

Many Thanks, i make a full backup than i try!

 

[LnxBil]

Many Thanks, i make a full backup than i try!

Very good, but for unsettings refreservation, you do not need to do that. You're just editing zfs metadata.

Do you have the "thin provisioning" check box activated in your PVE storage configuration?

 

[yena]

Very good, but for unsettings refreservation, you do not need to do that. You're just editing zfs metadata.

Do you have the "thin provisioning" check box activated in your PVE storage configuration?

No, is not flagged, i have noticed it only now! .. i think i have to flag it, right ?

 

[LnxBil]

Thin-Provisioning is a usual technique for virtualisation and helps with your problem.
You also need to reset the refreservation on each created dataset up to now:

https://forum.proxmox.com/threads/zfs-enable-thin-provisioning.41549/#post-200087