[Feature request] Option for auto remove older snapshots on lack free space about zfs-auto-snapshot HOT 9 OPEN

If --keep was required for this feature then I would withdraw my concerns.

from zfs-auto-snapshot.

This is a great idea. I would like to have this feature. How would free space be defined? Ie, space in the parent filesystem, headroom in the quota, ... ?

from zfs-auto-snapshot.

I think that easier way will be check free space for current filesystem, that becomes to be snapshoted. If this free space is shared with other filesystems, this will not the problem.

from zfs-auto-snapshot.

I would also like to see this feature added. It would be a great enhancement.

from zfs-auto-snapshot.

Explain more on how this would work. If --free=5GB it won't run if the pool doesn't have at least 5GB available?

from zfs-auto-snapshot.

I read the feature request as:

Given $ zfs-auto-snapshot ... --free=5GB

• Is free space less than 5GB?
  • If yes:
    • there is not enough space
    • remove oldest snapshot
    • go to beginning to check again
  • If no:
    • there is enough space
    • do snapshot

from zfs-auto-snapshot.

I think that might be a little … 'dangerous'. If we're so close to running out of space, starting to automatically delete (possibly big/huge) snapshots without the admin specifically deleting stuff he/she know is of no use might end up with all snapshots for ONE dataset/zvolume being destroyed, leaving all the other…

That doesn't sound like a solution I would want being done automatically...

from zfs-auto-snapshot.

The issue that I see with this feature is that you should also have a minimum number of snapshots to keep. Because, if you have 10 snapshots of a filesystem and you require --free=100GB but each snapshot is only taking up 5MB, then you will remove all of the snapshots and still not be able to take a new snapshot. That seems like a dangerous situation to me.

from zfs-auto-snapshot.

That's exactly why I don't think this should be done automatically...

from zfs-auto-snapshot.