Cluster migration NFS

[rf_pmx]

Currently I have a 3 node Proxmox cluster running VMs backed by qcow2 files on NFS.

As I'd like to reinstall the cluster with a new PVE version I thought about the following steps:

• make sure HA is not enabled on old cluster
• make sure no VMs are running on the old cluster
• remove one node from the old cluster (as outlined in the PVE documentation)
• reinstall the removed node with the new PVE version and create a new cluster with new cluster name
• add the NFS storage on the new one-node cluster with the same name as on the old cluster
• copy a test VM configuration to the new cluster
• start one VM for testing on the new cluster
• if that works, reinstall the two remaining nodes and join them to the new cluster

I know that mounting an NFS share on two distinct clusters is not supported since the assumptions about locking cannot be guaranteed across clusters. Also I'm aware that it's recommended migrating the VMs by backing them up and restoring them.

However, if HA is disabled on both clusters, no VMs are running on the old cluster and I make sure nobody touches the old cluster I assume my plan could work.
Or do I miss something?

(I've tested the procedure in a test environment and it worked just fine)

 

[czechsys]

Easy way - just disable nfs storage on old cluster.

 

[rf_pmx]

Good idea indeed!

 

[hec]

I would upgrade the cluster. Why installing a new cluster?

If you do not have some good reason i think this is wasting time and maybe some time for debugging to get all back running.

 

[rf_pmx]

@hec: I'd like to switch from legacy boot to UEFI, introduce a second corosync ring and skip one major version. While all of this can be done on existing systems, I think I'll be faster by just reinstalling the cluster nodes