Your machine’s Linux kernel panics. This is never something you’d want to see. Your entire server crashed over something. Much less pleasing is to find some worrisome log messages after the fact:
[479787.790225] kernel tried to execute NX-protected page - exploit attempt? (uid: 81) [479787.790247] BUG: unable to handle kernel paging request at ffffffffb585c2c0 [479787.790263] PGD b27a13067 P4D b27a13067 PUD b27a14063 PMD 8000000b27a000e3 [479787.790269] Oops: 0011 [#1] SMP PTI [479787.790285] CPU: 6 PID: 1118 Comm: dbus-daemon Kdump: loaded Tainted: G B I E X -------- - - <kernel version omitted>
Exploit attempt? Oh, really? Some code has gone rogue and tried to execute code from a memory page that was supposed to be non-executable for security purposes. What’s going on?
Hypothesis 1: Chinese hackers are at play! Nah, let’s be more realistic.
Hypothesis 2: Memory errors. Cosmic rays. It’s weird though. Usually memory errors are too subtle to notice, and to provoke this kind of error message is really something out of the ordinary. How unordinary? The server has 64 GB of non-ECC RAM (lacking any error correction). After all there’s a chance the memory just corrupted on its own.
The spoiler is in the title and after looking up the dates we had a perfect match. The server crashed on 22nd of January 2026, 23:47. In the late evening of the 19th we have been lucky to witness very intense northern lights over Germany. The solar storm had been hitting Earth between January 18th through 23rd.
Bingo. There’s your answer. Not just memory errors or mythic random cosmic rays from a galaxy far, far away. We were hit by actual cosmic “rays” from our sun. While we were having once in a lifetime experience – at least for non-hobbyists like me and my family – our server had been attacked by stellar forces to the point of crashing.
A post mortem? Use ECC memory. And maybe, just maybe, reboot your servers after solar flares are over? Even if ECC saved your CPU’s memory, are you sure your SSDs and other on-board RAM hadn’t suffered any bit flips? Error correcting filesystems like ZFS or BTRFS would be able to tell and fix your data before it’s too late (if coupled with ECC RAM, of course).