[mdlug] badblocks question

[Paul]

Jason Taylor <jmtaylor90 at gmail.com>, on Tue Apr 03, 2007 [01:57:25 PM] said:
> If the server hasn't had any recent dirty shutdowns, power fluctuations,
> etc. then I would lean to a hardware issue.
> 
> -Jason
> 
> On Tue, 2007-04-03 at 10:42 -0700, Dean Durant wrote:
> > Hello, I have run badblocks on some IBM eSeries blades
> > that I need to get fixed.   It reports about 2 million
> > bad blocks on an 80 GB disk.   IBM has me run their
> > diagnostics and they don't find anything.  I have run
> > fsck and it finds and fixes some filesystem issues,
> > but the bad blocks persist.   Is there anything else I
> > can do, and do people think it is a hardware issue, a
> > bad disk, or a filesystem issue?   Thanks, Dean

	Hi;

	smartctl is the tool you want to use to determine if
there are hardware problems with your disk. badblocks is a tool
more suited to disks from the ancient past. ext2/3 has legacy
support for mapping out bad blocks (see the -c option to e2fsck)
but a modern disk shouldnt have any bad sectors visible; the
drive should be mapping them out. In general if you see bad
sectors, the disk is toast. I cant vouch for your use of
badblocks, but it generally just tries to read chunks of disk,
and if it experiences an i/o error, deems the block bad. In that
case, I would expect your kernel logs to be overflowing with
these errors.
	Disk is cheap.

Paul
set at pobox.com