Proxmox 7.2 with 5.15.x kernel

[adamb]

As of late we have started having some odd crashes in our CentOS7 VM's which so far is only happening on various hosts that are on 7.2 running one of the 5.15.x kernels.

I have seen it happen on 5.15.35-1-pve and 5.15.39-1-pve.

Alot of the time it will happen after the VM is rebooted from within the OS, but I have yet to reproduce it myself in anyway.

All of our setups are HP Gen9/Gen10 severs with Alletra/Nimble iscsi based storage. All the 7.1 clusters have been solid.

Is anyone else seeing any issues on the 5.15.x kernels?

 

[sterzy]

Hi adamb,

do you see anything intresting in the logs (journalctl) either on the host or the VM when the crash occurs? What do you use for storage?

 

[adamb]

Hi adamb,

do you see anything intresting in the logs (journalctl) either on the host or the VM when the crash occurs? What do you use for storage?

The host side logs look to be pretty clean.

I will say we run a non standard mount option in our VM's with ext4 which is data=journal.

I partially suspect some odd issues with io_uring and using data=journal in the 5.15 kernel. We had some filesystem errors with io_uring and data=journal starting in 5.13, but they ended up being resolved in the testing repo's.

We run HPE Nimble & Alletra iSCSI storage.

 

[adamb]

I have also found that on Gen8 HP servers the only way I can get any 5.15.x kernel to boot is to disable VT-D in the bios. We have roughly 15 gen8's out in the field and so far ive had to do that to each one.

Ive been around Proxmox since 2.x, so far this kernel is not giving me a very warm fuzzy feeling.