Shutdown 3 nodes out of 5 nodes cluster in mix of v8.4 and v9.2

T

timlower

New Member
Oct 26, 2026
3
0
1
Hi
We have a 5 nodes cluster:
- 2 at v9.2.21 (upgraded recently) in physical DC1
- 1 at v8.4.21 (pending upgrade) in physical DC1
- 2 at v8.4.21 (pending upgrade) in physical DC2

As part of DC1 power maintenance, we need to shutdown the 3 hosts (VMs/CTs can shutdown).

Any advice how to safely power off that 3 hosts and then, power up later?

Thanks.
 
Hi @timlower, welcome to the forum.

Running mixed versions in one cluster isn’t supported as a long-term setup, and cluster-wide operations on a mixed cluster can behave unpredictably.

With all nodes on 9.2, you could use the cluster-wide HA disarm (ha-manager crm-command disarm-ha / arm-ha), which was introduced for exactly this kind of maintenance.
There is a manual workaround (pvecm expected), but it’s meant for recovery rather than planned maintenance, so I wouldn’t recommend it.

My suggestion: either finish the upgrade first and use the supported feature, or, if the maintenance can’t wait, build a small virtual replica of your cluster and rehearse the procedure there before touching production.

Cheers

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
 
bbgeek17 said:
Hi @timlower, welcome to the forum.

Running mixed versions in one cluster isn’t supported as a long-term setup, and cluster-wide operations on a mixed cluster can behave unpredictably.

With all nodes on 9.2, you could use the cluster-wide HA disarm (ha-manager crm-command disarm-ha / arm-ha), which was introduced for exactly this kind of maintenance.
There is a manual workaround (pvecm expected), but it’s meant for recovery rather than planned maintenance, so I wouldn’t recommend it.

My suggestion: either finish the upgrade first and use the supported feature, or, if the maintenance can’t wait, build a small virtual replica of your cluster and rehearse the procedure there before touching production.

Cheers

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
Click to expand...
Hi bbgeek17,

Thanks for your reply.

We have managed to upgrade all the five nodes to 9.2. So, in our maintenance scenario where we will need to bring down three nodes and keep the other two running with VMs and CTs. So, should we use 'freeze' or 'ignore' when running 'ha-manager crm-command disarm-ha' command?

Thanks again.

timlower.
 
Usually ignore is the most useful option as it will allow you to do anything with the VMs/LXC, freeze won't.
Worth mentioning which storage you use to remind you about Ceph MON quorum and PG distribution if that's what you use.
 
  • Like
Reactions: Onslow
You must log in or register to reply here.
Top Bottom
Back