[plug] Machine crashing (long sorry)

[Craig Ringer]

On Tue, 2004-02-03 at 17:07, Paul Arch wrote:

> So, before I go jumping to conclusions, does anyone on the list have any suggestions on how to pinpoint the problem ?  I thought reiserfs was stable.  Anyone else running with reiserfs ? I have been using it (on other machines) for many years without any problems.

The only time I've encountered this, it was due to severe filesystem
corruption from failing hard disks. I replaced the disk and it went away
for a while ... then that disk failed and I had a similar problem.
Hooray for that bad batch of Western Digital HDDs - it took me a while
to figure out that I had, in fact, just had two defective disks in a row
and that there was nothing wrong with the rest of the system.

In both cases, the FS was reiserfs. While extremely stable normally, I
have seen indications that it's less able to handle exceptional FS
corruption as well as ext2/3 - it tends to hang the kernel or force a
sudden reboot much more often than it should. Ext2/3 tend to oops
instead and leave the FS inaccessable (any app that tries to talk to it,
even df, will block forever).

So - check your disks. I posted a list of good ways to do this the other
day, but I'd start with using smartctl to query the drive's SMART data
and error log, using badblocks in read-only mode to scan for bad
sectors, and using the manufacturer's drive tools.

I would not be at all surprised to hear that your disk is failing and
reallocating bad sectors at a high rate, resulting in a corrupt FS.

DO NOT use reiserfsck - except maybe in read-only mode - until you're
sure the problem is not with the disk. It may only make things worse if
the underlying media is stuffed.

It's also worth noting that while the data partitions on the box I had a
similar problem with were resierfs, the system (/usr and /) partitions
were ext3. They were just as totally trashed, they just showed it a
little more gracefully.

-- 
Craig Ringer