Risks of upgrading Proxmox VE

[Mor H.]

We have a cluster with a bunch of nodes running the latest Proxmox v4 in production. The services we host on this cluster can't suffer prolonged downtime. A few minutes of downtime (say, for a reboot) is the most we can allow really.

We also have a Ceph cluster for storage of these VPS.

We would like to upgrade to Proxmox v5 and then Proxmox v6 but are extremely worried about the possible complications of the process.

It is our understanding that if things go wrong during the upgrade, we might end up having to restore our VPS from backups which will mean extensive and prolonged downtime.

Since it's not possible to upgrade from v4 to v6 directly, we would be doing it in two steps.

From your experience,
1. How risky is the upgrade? Are there high chances of things going wrong and extensive downtime happening?
2. What are the main things to keep in mind when upgrading to ensure it goes smoothly?
3. What is the most comprehensive and detailed guide you've seen for these upgrades?
4. What tests would you recommend we do before actually upgrading? Is it possible to somehow test the upgrade on 1 non-production node before doing it in production?
5. What are your recommendations to ensure that we have a solid contingency plan in case things don't go as planned?

Any and all help is appreciated!

Thank you.

 

[Vladimir Bulgaru]

Hey!

I've upgraded from Proxmox 5 to Proxmox 6 beta and it went very smooth. The update, in itself, is not that risky. What is risky though is the drivers support depending on the kernel version. Keep in mind that if you're running compiled drivers, DKMS or something driver dependent inside the cluster, i'd suggest testing it beforehand on a non-production instance.

I don't know if you are flexible with resources, but having a spare server would help a lot. You can either live-migrate or replicate instances across the cluster and avoid downtime if something goes wrong.

 

[oguz]

hi,

unfortunately upgrading from v4 -> v5 -> v6 would take some time and planning, downtime would be required since you need to reboot for the kernel upgrades.

during the update you will also be asked to restart some services so that might disturb your guests in some cases.

so if you are really worried about minimum downtime, then the best way to do the upgrade would be to install fresh 6.0 on a new machine and migrate your guests one-by-one by doing backup-restore on the new host (but you need to first set up your storages and network configuration etc. before starting the restore). that should minimize the frame of time where each guest stays down if you configure everything correctly.

edit:

check this: https://pve.proxmox.com/wiki/Bypassing_backup_and_restore_when_upgrading

you can do this also with ceph, if you connect your new 6.0 to old ceph (with same name for storage), you can copy your vm/ct configs and then boot them up from the new box while upgrading the old ones

[0]: https://pve.proxmox.com/wiki/Upgrade_from_4.x_to_5.0
[1]: https://pve.proxmox.com/wiki/Upgrade_from_5.x_to_6.0

 

[Mike Lupe]

Would this be a great moment to go for a Proxmox subscription?

We have a cluster with a bunch of nodes running the latest Proxmox v4 in production. The services we host on this cluster can't suffer prolonged downtime. A few minutes of downtime (say, for a reboot) is the most we can allow really.

We also have a Ceph cluster for storage of these VPS.

We would like to upgrade to Proxmox v5 and then Proxmox v6 but are extremely worried about the possible complications of the process.

It is our understanding that if things go wrong during the upgrade, we might end up having to restore our VPS from backups which will mean extensive and prolonged downtime.

Since it's not possible to upgrade from v4 to v6 directly, we would be doing it in two steps.

From your experience,
1. How risky is the upgrade? Are there high chances of things going wrong and extensive downtime happening?
2. What are the main things to keep in mind when upgrading to ensure it goes smoothly?
3. What is the most comprehensive and detailed guide you've seen for these upgrades?
4. What tests would you recommend we do before actually upgrading? Is it possible to somehow test the upgrade on 1 non-production node before doing it in production?
5. What are your recommendations to ensure that we have a solid contingency plan in case things don't go as planned?

Any and all help is appreciated!

Thank you.

 

[LnxBil]

Would this be a great moment to go for a Proxmox subscription?

Every moment is a great moment to go for a Proxmox VE subscription :-D