2-Node Ceph Cluster issues during graceful shutdown of 1

[cyrus104]

For my homelab, I have 2-nodes running PVE 8.1.10 with ceph installed. Both have monitors,mds,mgrs, and osd. When everything is up it's working good. When I do an update and reboot either of the nodes, the other node soon becomes unresponsive for a few minutes and the VMs running on that note also shutdown even if they are running on a local-lvm and not the ceph.

Where would I check for logs on why this happens and possible ways to fix it.

 

[leesteken]

Two-node cluster are always problematic (since each node does not have quorum (more than half of the (two) votes) on their one). Ceph needs at least three nodes (without redundancy). Please read the manual and search the forum as this is a very common mistake.

 

[cyrus104]

Thanks for the quick response, I thought with increasing one of the nodes to two votes that it would be "stable". To confirm Ceph needing three nodes is a separate issues than the cluster quorum issue?

Thanks

 

[LnxBil]

To confirm Ceph needing three nodes is a separate issues than the cluster quorum issue?

Yes, both technologies need at least 3 nodes, so if you only want to run with two, you need to change the behaviour of both technolgies to force them to work in a non-supported solution that is prone to fail.