Hi. On Tue, Jun 08, 2021 at 04:44:35PM +0500, Alexander V. Makartsev wrote: > > How can I debug this problem? My suspicion is that this has to do with the > > kernel upgrade between stable and testing (4.19 to 5.10), but I'm not sure. > > The way I would approach this problem is by making boot logs > persistent [1] so they could be searched for clues (kernel module > name, device hint, etc) after each boot.
I'd like to suggest a different approach, considering we're dealing with kernel panics here. journald is merely a userspace program, so it cannot process kernel panics reliably (barring kernel OOPSes, of course) - because if you have a kernel panic, you cannot write in any filesystem. They've invented kdump - [1], [2] with exact purpose of capturing kernel panics and storing kernel crash dumps in a persistent way, and crash - [3] - for analyzing them. Of course, in its current form kdump requires a real hardware, but it's hardly an issue here. Reco [1] https://mudongliang.github.io/2018/07/02/debian-enable-kernel-dump.html [2] https://packages.debian.org/buster/kdump-tools [3] https://packages.debian.org/buster/crash