Separate a Node Without Reinstalling

[RolandK]

Is this still valid/up2date with current proxmox ?

https://pve.proxmox.com/wiki/Cluster_Manager#pvecm_separate_node_without_reinstall

i need to eject a node from a cluster and keep the VMs.

can this be done without vm downtime, i.e. can i leave the vm's up and running or do we need a maintencance window ?

just asking, because i cannot afford hassle with that system.

 

[t.lamprecht]

Is this still valid/up2date with current proxmox ?

Yes, it's up-to-date, but it's also mostly targeting installations that are relative vanilla w.r.t. clustering.

i need to eject a node from a cluster and keep the VMs.

How many nodes are there in the cluster?

can this be done without vm downtime, i.e. can i leave the vm's up and running or do we need a maintencance window ?

If the VMs have no resource tied to that node, then it should work out fine.

If you use HA, it might be best to disable that first. In essence that means stopping the pve-ha-lrm.service first on all nodes and once all LRMs are indeed stopped then also stop all pve-ha-crm.services (ideally the node that is the active master last to avoid some churn). Once you're done, you can start the HA service in reverse order again (first CRM then LRM).

Do you use ceph, or some other shared storage that this node participates in? Then it should be cleaned up first.

just asking, because i cannot afford hassle with that system.

Depending on how critical this is, you might want to play it through with a virtual PVE three-node cluster first. I.e., install PVE in VMs or test servers, cluster them and create a few (dummy) guests for a more real life scenario, you can then even create a snapshot of the PVE VMs on the underlying host level to allow repeated experiments, and then follow the docs for separating them.

 

[RolandK]

thank you.

the cluster will not remain afterwards, all three cluster nodes should be consolidated to the first node. the first node has already been replaced by putting the disks into a larger machine with enough ram to handle all machines from all three nodes, and all VMs already have been moved to the first node.

i just want to get rid of the clustering and the two empty nodes

 

[RolandK]

worked like a charm - and without host or VM downtime

thanks for help.