With the deprecation of PVE 8 coming in August, this is for general informational purposes in case something similar arises again with other kernel builds.

Running PVE 8 with up to kernel 6.8.12-8 on two servers (AMD Epyc v4 processors) the systems were stable for months of uptime throughout 2025 with no reboots other than planned ones.

When one server was upgraded and received kernel 6.8.12-19-pve+ this spring it began exhibiting random reboots one or more times per day. The boot logs indicated a hardware error on CPU 2.

When the second server was upgraded to kernel 6.8.12-29-pve it immediately began exhibiting the same random reboot behavior and an identical error message.

We then upgraded the first server to PVE 9.2.3, with kernel 7.0.6-2-pve. The random reboots ceased and system logs contained no indication of the hardware error. After observing the first server for a few days the second server also was upgraded to PVE 9.2.3 and it became stable again.

The hypothesis is that something about 6.8.12-19-pve and later 6.8.12 builds had issues with the ASUS dual socket server chassis and Epyc v4 processors.