[Gllug] Ext3 problems

[Jan Kokoska]

Just as Martin suspects, I think its the disk dying.

Using smartsuite or smartmontools, view the log by:
'smartctl -a /dev/hdX'

Non-zero reallocated sector count is not good (drive controller
reallocates space in bad sectors from the factory dedicated area, but
you don't really want to exploit this feature in regular use). 

You will most likely have something in error log, but while empty log is
a good thing, non-empty log doesn't mean anything. Tampering with DMA
modes or HPA (host protected area) and other non-standard stuff might
fill it, while the hdd is still in for years of reliable service. If you
can decrypt the log further, I for one would like to hear about it.
Depending on the package, test triggers for smartctl vary, see man pages
to run the S.M.A.R.T. tests. 

If it bothers you, do the simple thing. Tar it all away (if you don't
have backup yet, which would be a shame) and sequentially rewrite the
whole hdd, then read it back.

I use: 'dd if=/dev/zero of=/dev/hdX bs=1M conv=sync; dd if=/dev/hdX
of=/dev/null bs=1M'

Watch the beast in another console by 'vmstat 1' and if it bombs out
with DMA or whatever errors or stops at one place (IO idle, disk
spinning, LED lit up), you know where the bin is...

Disks are cheap these days, much cheaper than your data anyway.

Jan

P.S. RAID level 1 is your friend when dealing with IDE drives, at least
Linux SW if not proper 3ware.


On Tue, 2004-09-21 at 14:38 +0100, Dylan wrote:
> Hi All,
> 
> I've been trying to debug a problem for a friend... Three days ago, he 
> found that there were suddenly permission problems and missing files in 
> his home directory (on /dev/hdb2). He rebooted, and got dropped at a 
> runlevel 1 login after fsck failed. I had him fsck the fs manually, and 
> all seemed fine after the subsequent reboot. 


-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug