[plug] server failing with bizarre disk errors

[Craig Ringer]

> Old ATI chipset with a Celeron 400 from when it was the best thing any home
> user could contemplate having.

... except a Pentium. That said, the older celerons were far less 
crippled than the new P4 celery.

>  hda is 20GB Seagate, and hdb is a 13GB
> "Bigfoot" - which I think was a WD drive. Problems appear to start on hdb,
> except that my main distro (SuSE) resides there, and it (almost) never has
> these problems. In fact checked the logs last night and there was nothing
> for the past 2 weeks - even though it'd been on 90% of the time, running
> SETI at home. (OK, not particularly disk intensive - but it didn't whine when
> I did a few things before shutdown, and the "updatedb" jobs ran perfectly
> every night.)

I have found the Quantum BigFoot drives to be reliable storage, if you 
can live with the quirks. They don't like some BIOSs and other hard 
disks, tending to do funny things like misreport geometry on boot. 
They're also about the slowest harddisk in the universe.

I don't get logs of the disk errors - its like all ATA writes /stop/ 
when the errors ocurr.

Have you enabled S.M.A.R.T and queried your drives (if possible)?

I've just moved the drives on my problem server to separate interfaces, 
clocked down the CPU, disabled the parallel port and clocked down the 
RAM. It /seems/ to be behaving now, but it'll be hard to say for sure 
for a few days.

The western digital drive tools reported "drive OK" but I don't really 
trust them - SMART says 54 reallocated sectors, that's not OK. Then 
again, vendor-specific SMART data can be hard to interpret correctly.