Inplace Upgrade of 3.1 to 4.X Cluster

[Jospeh Huber]

Hi,

I have a 5 Node production cluster (with local) storage.
So i have already read the wiki entry about that.
https://pve.proxmox.com/wiki/Upgrade_from_3.x_to_4.0

I have already done an in-Place testupgrade with a seperate machine ... without cluster... works great.

My Question is on this part:
It is not possible to mix Proxmox VE 3.x and earlier with Proxmox VE 4.0 cluster
Due to the new corosync 2.x, the cluster has to be re-established again. Please use the same clustername.

My plan is, to migrate node by node. But there are so much vms, that i cannot migrate it overnight.
Is it possible i.e. to upgrade the first Node, create the cluster again (with the same cluster-name) and the old cluster with the same cluster-name and the old corosync can operate at the same time in the same network without any Problem?
Use another Cluster-name?
What is the best strategy to to such a longtime migration?

Thx.

 

[t.lamprecht]

Is it possible i.e. to upgrade the first Node, create the cluster again (with the same cluster-name) and the old cluster with the same cluster-name and the old corosync can operate at the same time in the same network without any Problem?

Yes, this is possible. Be sure to remove the upgraded node from the old cluster configuration (after! the upgrade), so that no conflicts happen.

pvecm delnode <already-upgraded-node>

(this only deletes the node from the corosync config, no other data (e.g. the deleted nodes cfg files) gets touched/deleted)

It is not possible to mix Proxmox VE 3.x and earlier with Proxmox VE 4.0 cluster

This should mean that its not possible to have a cluster with some 3.4 and some 4.X installations, as they do not see each other.
I have various clusters on my network (3.4, 4.x) without any problem.

Use another Cluster-name?

Not needed.

Just be sure that you follow the guide precisely and do not miss a step, then you can upgrade one node after the other, delete it from the old cluster and add it to the new one.
If you lose quorum on one node due the fact that its the only one left you can achieve it temporary again through setting the nodes expected vote count to 1:

pvecm expected 1

In general I really would recommend that you test a full upgrade on a test machine (or VM with PVE installed).

 

[Jospeh Huber]

Hi Thomas,

thanks for your quick and good answer!

I have already done a complete Migration 3.x => 4.x following the steps from a separate machine outside the cluster (inlcude openvz migration) ... that works great.