Recovering replicated VMs on non-HA cluster?

[hefferbub]

I have a 2 node cluster running 6.3 where I have a primary server than normally runs all the VMs and a smaller backup server intended to be ready to run those VMs in case of planned or unplanned downtime on the primary.

I do a regular replication (every 2 hours) of all critical VMs to the backup server.

For planned downtimes, everything is simple. I can simply manually migrate all VMs to the backup, which is fast since they are mostly up-to-date already because of regular replications.

What I'm not clear on is what will happen if i get a crash on the Primary and need to bring up the VMs on the backup?

I'm aware that those VMs will only be current as of the last replication, and I'm also aware that I will temporarily have a no-quorum cluster, so presumably I will have to issue a

pvecm expected 1

.

But beyond that, what needs to happen to launch these VMs on the backup?

Thanks!

 

[Stoiko Ivanov]

Are you using storage replication for the sync?

In any case - recovering a VM to another node in case of failure is described in that chapter:
https://pve.proxmox.com/pve-docs/chapter-pvesr.html#_migrating_a_guest_in_case_of_error

keep in mind that doing cluster-operation when not running with quorum can cause data-loss (if the main node comes back up and still considers itself the owner of the VM, and you also run `pvecm expected 1` there for example).

Consider adding an external vote (via qdevice):
https://pve.proxmox.com/pve-docs/chapter-pvecm.html#_corosync_external_vote_support

and read the warnings regarding running without quorum in the reference documentation.

I hope this helps!

 

[hefferbub]

Thank you so much. I thought I had seen documentation on that, but could not find it.