Day: November 15, 2004

After last night’s hang, I resolved to do some testing to see if I need a new motherboard or something.

I booted my computer with a non-smp kernel into single user mode. This would eliminate most “distractions” or other causes for the error. Mounted /dev/hdc1, which is an ext3 file system. I did a simple tar of some of the files on /dev/hdc1, and after about 30 seconds it froze up.

So then I booted with a Mandrake Move CD, which has a 2.4 kernel. I mounted /dev/hdc1 again, and did the tar and it worked. Then I mounted /dev/hdc4, my mp3 collection, and did something that’s frozen up my computer every time since I upgraded:


find . -type f -print | sort > /tmp/mp3.files


It didn’t freeze. Not only that, but my mp3s appear to still be there (although the presence of large numbers of files in /mp3s/lost+found makes me think that some of them will be pooched.)

So now my question becomes – how do I get back to a 2.4 (or maybe a 2.6 non-Fedora) system without incurring another 2 days of downtime and headaches? I won’t be around this weekend, and I don’t particularly want to risk this happening every night for two weeks.

I woke up this morning to find every screen logged into the server was showing “unable to contact the UPS” errors. One window was still responding a bit, and an “uptime” command showed the load average just a hair over 230, and rising.

After power cycling, I find the log shows that sometime around 2am, when the nightly cron jobs kick off, the second IDE controller started throwing errors again.

I’ve got to consider the following possibilities:

• The hardware just miraculously decided to fail when I upgraded.
• The hardware was always a little bit bad, but the 2.6 kernel notices the problem and the 2.4 kernel didn’t. OR
• There is nothing wrong with the hardware and it’s a fault in the kernel

Tonight I’m going to have to go offline again, while I try booting with a Knoppix CD with a 2.4 kernel to test the hardware again. If that works, then I’m going to try 2.6 with no smp, and with the infamous “noapic” flag (whatever the hell that means).