Proxmox cluster node upgrade process with minimum downtime

[damiengm]

Hi
Just want to confirm the node upgrade for a HA cluster. Setup is a 2 node cluster with q-device. Storage is shared RAW on iSCSI NAS. VMs ideally not to be shutdown. is it:
1) live migrate running VM's to node B.
2) apt-get update (or gui upgrade) node A (and reboot)
3) live migrate running VM's to node A
4) apt-get update (or gui upgrade) node B (and reboot)
5) update q-device
6) redistribute VM's as required

current version of Proxmox: 9.2.4, so this a patch update.

Do I need to do anything in tha HA section? any other things to look out for or missing steps? Thanks

 

[bbgeek17]

Hi @damiengm,

There are several different layers that can be upgraded. Depending on the upgrade, you may need to update just one of them, a combination, or all of them.


1. PVE packages (GUI, CLI, etc.)

These can be upgraded without migrating or rebooting any VMs. The upgrade will restart PVE services as needed, without impacting running VMs.

2. Core PVE packages (e.g., QEMU)

QEMU is what runs your VMs. Even when upgrading these packages, you don't need to reboot the host or migrate your VMs.

The new binaries are installed on the system, and any VMs started after the upgrade (i.e., cold-started) will use the new version. VMs that are already running will continue using the old binaries until they are migrated or cold-started.

Whether you want to include this as part of the upgrade or defer it to a later maintenance window depends on your business requirements.

3. OS kernel

The kernel upgrade is similar in that the new kernel is installed but does not become active until the host is rebooted. Again, whether you need to do this immediately or can defer it to another maintenance window is up to you.

You can even run a cluster with mixed software versions, although this is not recommended as a long-term production configuration, and certainly not across major versions.

In short, it all depends on your business requirements and maintenance constraints.


The process you described is one way to approach the upgrade. I'd also look into the newer Maintenance Mode feature, which allows you to suspend HA decisions during maintenance. It's covered in the PVE Cluster Administration Guide.

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

 

[damiengm]

Hi @bbgeek17
Looking at the pending updates it looks like everything has an update. I think I will use the Maintenance Mode feature, used that feature in VMware for multi host updating.

 

[SteveITS]

If you enable/disable maintenance mode PCE will migrate VMs for you. And migrate them back again.

The QDevice can be updated at any time.

If using the command line, don’t use apt upgrade, per the docs.