Snapshotting guests on shared replicated storage?

D

dswartz

Renowned Member
Dec 13, 2010
286
9
83
So I am running two PE2.3 hosts in a cluster using drbd as shared local storage. Works perfectly. My only complaint is that I can't snapshot any of the guests, since the storage is not snappable (lvs on the vg on top of the drbd devices.) What is frustrating is that since the storage is lvm-based, there is no inherent reason snapshots can be done, is there? I had been used to being able to snap a guest, update a software package and be able to roll back if it failed horribly. This worked under vmware esxi and xenserver, but I can't do it here unless I use (apparently) qcow2 vm disks on nfs?
 
dswartz said:
So I am running two PE2.3 hosts in a cluster using drbd as shared local storage. Works perfectly.
Click to expand...

DRBD is not shared storage, its like a raid1 over the network. If you use DRBD you should be aware of what you are doing.

dswartz said:
My only complaint is that I can't snapshot any of the guests, since the storage is not snappable (lvs on the vg on top of the drbd devices.) What is frustrating is that since the storage is lvm-based, there is no inherent reason snapshots can be done, is there? I had been used to being able to snap a guest, update a software package and be able to roll back if it failed horribly. This worked under vmware esxi and xenserver, but I can't do it here unless I use (apparently) qcow2 vm disks on nfs?
Click to expand...

Since 2.3, you can do backups without the use of LVM. see http://pve.proxmox.com/wiki/Backup_and_Restore
If you want to use snapshots like described here, see http://pve.proxmox.com/wiki/Live_Snapshots - DRBD does not support snapshot here.
 
Let me clarify. I am using the replicated local storage in shared mode. This is documented and present in your own GUI. So it sounds like until some future enhancement, whether I continue with drbd or switch to rbd or somesuch, snapshots of running VMs will not work...
 
yes. backup works in your setup, snapshots not.

you mentioned it worked under vmware or citrix xen - I assume you did not use DRBD or any other replicated storage there?
 
No, in both cases it was the standard file format (e.g. no choice between qcow2 or raw.) I would use qcow2 here if I could but I understand why it has to be raw format stored on an LV.
 
if you use block devices directly like lvm volumes, there is no virtual disk format.
 
Semantics. Obviously the layout on storage has a format. e.g. vm disk metadata that qcow2 has to support snapshots etc. That has nothing to do with the file itself, but data inside it. That said, like I said, I understand why it works this way - it's just a nuisance that I can't use LVM snapshots instead of file snapshots. I don't even care if it didn't allow snapping running VM state - just the ability to do a 'power cord rollback' would be fine. AFAICT, I can do this from the CLI anyway, no?
 
You must log in or register to reply here.
Top Bottom