Live migration fails (freezing, kernel panic, etc..) from newer CPU to old CPU only; using latest PM 7.3-4

[mailinglists]

Hi guys.

I have upgraded my cluster from PM6 to PM7.3-4. Versioning: pve-manager/7.3-4/d69b70d4 and Linux 5.15.83-1-pve.
I observed that live migration does work only from older CPUs (like E5-2690 v2) to new ones (like Gold 6154) but not the other way around.
When migration fails, the VM instance becomes unresponsive in all aspects. Sometimes we get kernel panic, sometimes CPU hangs. I have attached some screencaps below.

How to debug this or even better, how can we solve this?
I saw some threads with similar issues, that suggest upgrading the kernel?
What do PM devs think?

 

[Neobin]

I saw some threads with similar issues, that suggest upgrading the kernel?
What do PM devs think?

Then you should also have seen, that the developers, not only once, already suggest to try/use the 5.19 opt-in kernel: [1] on the source and target node; e.g.: [2].

There exists also the recent 6.1 opt-in kernel: [3].

[1] https://forum.proxmox.com/threads/opt-in-linux-5-19-kernel-for-proxmox-ve-7-x-available.115090
[2] https://forum.proxmox.com/threads/l...-to-lower-frequencies-cpus.120766/post-524779
[3] https://forum.proxmox.com/threads/opt-in-linux-6-1-kernel-for-proxmox-ve-7-x-available.119483

 

[mailinglists]

Thank you for your answer. Because things change with time, I thought it was best to ask.

So the official response to solving this bug would be, upgrade to 5.19 kernel. We will do that and report back only if any issues arise.

 

[mailinglists]

Those threads also stated I should update all nodes to newer kernel for live migration to work, but I just updated the newer (Intel scalable) servers to: Linux 6.1.10-1-pve #1 SMP PREEMPT_DYNAMIC PVE 6.1.10-1 (2023-02-07T00:00Z) and live migration works in all directions now.

So unless devs say, I should upgrade all other nodes, seems I will update to 6.1* just the newer CPU ones.