pve zfs and drbd
- Thread starter RobFantini
- Start date
*bump*I'm looking at a two node setup with ZFS and DRBD. Cheapest way for redundancy. I saw a guide on how to do it, but it was before ProxMox had official ZFS support, so it's probably not accurate anymore.
Hi
Yes, I'm using drbd ontop of zfs (zvols) on a two node cluster without issues.I had zfs installed before proxmox integrated it in the latest versions.I've updated proxmox to latest version and I'm using that zfs version now.I'm using separate drbd resources per vm (zvol).
Yes, I'm using drbd ontop of zfs (zvols) on a two node cluster without issues.I had zfs installed before proxmox integrated it in the latest versions.I've updated proxmox to latest version and I'm using that zfs version now.I'm using separate drbd resources per vm (zvol).
Hello acidrop, Are the drbd resources primary/primary or primary/secondary?
Last edited:
in order to use the drbd resource from storage > add > zfs , which steps would you follow?
Actually I have created zvols before zfs gui integration in proxmox, so I created them manually by using zfs commands to create zvols.Each zvol same size on both nodes.After creating zvols use drbd config to attach those zvols to separate resources(which corresponds to each vm disk).After doing that create the apropriate PV and VG on drbd resources as described in wiki.Finally attach VG via storage -> lvm to proxmox.
It sounds complicated but once you do this some times you will get more familiar.
It sounds complicated but once you do this some times you will get more familiar.
that sounds straight forward . I used zfs some time ago and will be glad to use it again. I think pve / zfs / drbd is a great combo . no hurry can you explain more of how you use and manage zfs cli snaphots ? Do you backup the snapshots to another system?
I use zfs autosnapshot script (https://github.com/zfsonlinux/zfs-auto-snapshot) which takes automatic hourly,daily,weekly,monthly snapshots on each host.
I use also zfs send/recv manually to copy this zvols on a external storage over network.The only drawback is that in case of snapshot rollback you have to bring down vm,drbd resource, then do the rollback and after bring them back again.
I use also zfs send/recv manually to copy this zvols on a external storage over network.The only drawback is that in case of snapshot rollback you have to bring down vm,drbd resource, then do the rollback and after bring them back again.
if the main server room burnt down and had to be replaced using off site receive target, how would one get the kvm up and running again?
You could have a third proxmox node offsite,with similar configuration as main site, which will receive zvols(an their snapshots) via zfs send/recv.Not sure about this since I haven't tried it.
The main thing is that you won't loose your data since your zvols will be intact.
hello acidrop, thank you for the responses. our off site backup servers do run pve , will convert to pve + zfs and also figure out how to use the received zvols later . I think it'll be fairly easy to do .
Have you tried using openvz on a zvol? I tried that sometime ago and it did not work out the way it was done.
I just tried following parts of http://zfsonlinux.org/example-zvol.html to make a ext2 zvol, adding directory to pve, restored a vz backup. the vz would not complete startup.. it was stuck trying to run ntpdate . So I'll stick with kvm on zvol testing.
I am not sure I know what you meant about 'native zfs dataset' . Is that just using filesystem made at pve install?