2 nodes 2.x production cluster: how to test 3.1 on one node?

[m.ardito]

Hi,

I have an outdated 2-nodes 2.x cluster (yes I know 3 is recommended) in production.

I wish to test the new 3.1 on one node and test vm using backups. I need to test the same hardware I will use in production, so I can't use another test machine.

Once I am fine with hw tests, I can convert both nodes to 3.1 to get a new 2-nodes 3.1 cluster. And I also got basic subscriptions approved by the management ;-)

what is the best approach I can do?

I thought
1) move all vm to node1
2) remove node2 from cluster
3) update node2 to 3.1
4) test vms on node2 (from backups)

and, when all goes well then either

5a) backup/restore all vms from node1 (2.x) to node2 (3.1)
6a) upgrade node1 to 3.1
7a) create a new cluster on node2 (which has all vms)
8a) join node1 to new cluster

5b) stop all vms on node1
6b) upgrade node1 to 3.1 (which has all vms)
7b) expect to have previous cluster on node1
8b) join node2 to previous cluster

what you suggest, is there any better/safe path hopefully keeping production stop time to a minimum?

Thanks,
Marco

 

[stef1777]

If you have servers with remote console access. It will be may be better to install 3.1 from scratch than upgrade. Faster and more reliable.

 

[m.ardito]

Ok, I could reinstall or upgrade, but my question was about the overall process:

best way to "detach" one node to test 3.1 on it and, when all is tested, upgrade the whole system to the 3.1 version, with minimal risks/impact on vm activity.

Anyone can share experiences?

Thanks,
Marco