Leave a cluster ... am I doing something wrong?

[tcabernoch]

When you remove a PVE host from a cluster, it seems to be pretty b0rked after that.

The cluster itself requires the pvecm remove thing and also delete the /etc/pve/nodes/nodename and also manage the ssh keys. Hope ya didn't have HA configured. It's messy, but after enough cleaning, you can get the cluster to forget about the missing host.

But the host itself seems to think all sorts of things that aren't true. It seems like its just broken. Like you'd be better off formatting the drive.

Are there steps to restore a PVE host that has been removed from a cluster back to full function as a standalone?

 

[esi_y]

Leaving a cluster is of course cardinal sin of any PVE user.

When you remove a PVE host from a cluster, it seems to be pretty b0rked after that.

And Proxmox try to warn against it in their docs [1] - it is somewhere there in a vague way, it mentions some potential for leftovers.

The cluster itself requires the pvecm remove thing and also delete the /etc/pve/nodes/nodename and also manage the ssh keys.

Yes they do not care for this, I once asked, was told that auto-deleting might be beyond user's expectations, i.e. it is safer to keep it there. When I asked why it's not moved to e.g. deleted/ subdirectory, no follow up.

Hope ya didn't have HA configured. It's messy, but after enough cleaning, you can get the cluster to forget about the missing host.

But the host itself seems to think all sorts of things that aren't true. It seems like its just broken. Like you'd be better off formatting the drive.

This is definitely the so-called supported way.

Are there steps to restore a PVE host that has been removed from a cluster back to full function as a standalone?

Basically (on the removed node):

systemctl stop pve-cluster
systemctl stop corosync
pmxcfs -l
rm /etc/pve/corosync.conf
rm -r /etc/corosync/*
rm -r /var/lib/corosync/*
reboot

Then you still have to cleanup the configured leftovers in /etc/pve/ that do not apply anymore, you found out about those already.

[1] https://pve.proxmox.com/wiki/Cluster_Manager

 

[tcabernoch]

Thank you.

I've learned well enough not to try this at work, but I want to join a physical host to my homelab cluster, migrate or clone a couple VMs, and then leave the cluster again ... and not break my standalone host. Of course, this is something you can do with the VMware suite, back and forth, all day long, without issue. We are not using VMware.

This is good stuff. I might try it with a virtual host before I risk my physical server.

 

[esi_y]

if that’s your only reason, are you aware of qm remote-migrate?

 

[tcabernoch]

No. But I'm aware of pvezsync. Same thing. Why should I have to CLI something so simple?
And that's just one use case. Perhaps is a poor example. Joining and leaving clusters is the point of my inquiry.