,
10 tweets,
2 min read
Read on Twitter
My home Linux machine is definitely unreliable now, but it's not clear if it's hardware or software. I'm hoping software, but the symptoms sort of look like it's powering off sometimes and not coming back for ... a while. (But sensors show no abnormal readings just before.)
Unfortunately this is where I really need a second machine at home to capture any kernel netconsole information and so on, regardless of how small such a machine might be. Perhaps it's time to finally get a Raspberry Pi.
At this point, either my home machine is taking ~6 hours between kernel hang and reboot or it's spontaneously powering off and then coming back on again later. Sadly I suspect it's the latter. PSU? Who knows.
For various reasons, I'm cautiously optimistic that my home machine's instability problems are due to kernel panics, not hardware. I even have a suspect in mind; I may have blown my own foot off with ZFS ARC memory settings.
My cautious optimism about my home machine is back to pessimism, because the symptoms I can see don't really match my theory for a kernel panic. If only my desktop would crash or lock up in front of me, so I could see the failure in person.
My latest crazy theory about my home machine's ongoing hang/reboot problem is that it's triggered by the ambient air temperature getting too low (which suggests a PSU issue for various reasons). If it wasn't snowing madly in Toronto right now, I'd test that the obvious way.
My home machine doesn't seem to be losing power when it locks up, so my latest suspect is the CPU fan going too slowly when the ambient environment gets too cold. I've set an explicit minimum fan RPM, so we'll see how that goes.
So I definitely appear to have the most ironic of things, a computer that doesn't like it when it's too cold. I don't think it's the CPU fan or the CPU, either, which leaves me strongly suspecting the motherboard. I guess it's time to update to the latest BIOS.
My home computer turns out to be less than a year old, so this may really be the first time it's been in an environment this cold (I built it at the end of March 2018 and Toronto is usually warming up by then).
My vague attempts to re-seat the only really re-seatable bits of my home computer have now demonstrably failed to stop it locking up when things get too cold, so I guess it's time for a new BIOS and maybe a blog entry.
