Zfs vms in dataset

[redmop]

Is there a way to set up the zfs plugin to store vms under a dataset instead of the root of the pool? This would make scripting snapshots and send/recv much easier.

I other words, instead of
Rpool/vm-100-disk-1
Have
Rpool/pve/images/100/disk-1
Or even
Rpool/pve/images/vm-100-disk-1

 

[vkhera]

I filed a feature enhancement request for this too: https://bugzilla.proxmox.com/show_bug.cgi?id=703

I would like to be able to set ZFS tuning per VM this way and not have such a cluttered top level pool.​

 

[dietmar]

You can already use subvolumes, for example

zfspool: mysubpool
    pool testpool/mysubpool

 

[vkhera]

Oh nice! It would be great if the "add storage" UI would let you pick an existing data set rather than only offering the top level pools to choose from.

 

[dietmar]

Oh nice! It would be great if the "add storage" UI would let you pick an existing data set rather than only offering the top level pools to choose from.

That is already implemented in PVE4.0 ...

 

[redmop]

Awesome. Does this work with lxc too?

 

[dietmar]

Does this work with lxc too?

yes

 

[redmop]

Just tried it out. Works awesome. Y'all rock!

 

[vkhera]

The next improvement would be to have each VM make its own (sub) data set to hold the disks, like how the directory based storage makes a subdirectory per VM. This would allow for some useful ZFS management.

 

[redmop]

Good idea.