[SOLVED] 6.2 No HA - Replication plus migration when the active node goes down?

[hefferbub]

I love the replication feature with non-shared ZFS storage! It is a thing of beauty to watch these rapid offline migrations which only need to transfer a modest differential of data. It is just what I hoped for to allow for quick failover I manage myself (and avoid the sometimes-arcane complexities of HA).

However, it seems like the beauty fails when the node that hosts the VM goes down and migration is no longer possible.

Am I missing something? Are things operating as designed or have I made a mistake?

In my test environment I have a VM running on node 1 and being replicated to node 2 (without enabling HA on the VM). When both nodes are running it works great and I can migrate it back and forth easily. However, when it is running on node 1and I shut down node 1, I can no longer do the migration to node 2; it just says "no route to host".

I would have expected that the cluster would know that node 1 is offline and would use the corosynced config info to let me bring it up on node 2.

I can understand that node 2 would only reflect the state of the VM since the last scheduled replication, so doing this could entail some data loss. But what is the thinking of the developers for people in this predicament?

Is there some kind of manual process to follow in the case that one chooses to bring up the replicated VM on node 2?

Thanks. And thanks for this amazing product!

 

[wolfgang]

Hi

Is there some kind of manual process to follow in the case that one chooses to bring up the replicated VM on node 2?

Yes, you must steal the config manually, but this is easy.

cd /etc/pve
mv /nodes/<Node1>/qemu-server/<VMID>.conf /nodes/<Node2>/qemu-server/
qm start <vmid>
[Code]