Folks whats the best logs to check to try and work out what my server is doing before it locks up? Its been solid as a rock for ages with months of up time now its freezing every few hours, i've not installed any new software beyond updates. Its a dual p3 running etch.
If it's cyclic, I'd check out if by some reason ACPI(d) isn't installed, or renabled in the BIOS or something. As for logs, /var/log/syslog will be your best bet, paired with /var/log/messages.log
I've stopped the ACPId on your advice tbh i though i had it running before maybe something has changed and its causing a bug. Although i thought that was the point in Debian stable, nothing changes beyond security patches. There doesn't appear to be any pattern in the logs as to what the previous few events are before it logs the restart. I'll leave acpi off for a few days if it doesn't crash then at least i've found the problem. Changed my mind, pulled it out of the cupboard and hooked it up to a screen and ran a kernel compilation on it: Kernel panic cpu context corrupt. Bugger looks like something on this old beast is failing (failed).
Oh, it's not booting? I thought it was. "dmesg" gives you all info on hardware events, look in there.
It boots fine, it appears to intermittently lock up under little load and lock up pretty quickly when its doing some thing intensive. I'm leaning towards the memory, but to be sure i passed noacpi to the kernel on boot and i'm trying the kernel compile again. The compiling the kernel is just the best system stress test i can think off right now, besides i actually do need an updated kenel for another box so if it does work bonus it saves me a job later.
You run a custom kernel on a Debian box? To check the memory there's memtest (same as a lot recommend on Windows). To get cpu usage to 100%, there's an easy trick Code: cat /dev/urandom >> /dev/null
Nope stock kernel on the Debian server, custom kernel on newer machine to get some bits working which pre 2.6.20 kernels don't support. Etch is still on 2.6.18. Yeah just about to try memtest, i've got cpu burn in for checking the cpu, its good because it will tell you if there were errors which did not crash the machine. Just out of interest why do you ask if i run a custom kernel on Debian?
