FreeBSD 8.1-R/amd64 - zfs 'hangs' - help tracing?

[Jeremy Chadwick]

On Thu, Sep 30, 2010 at 09:08:39AM +0100, Karl Pielorz wrote:
> --On 29 September 2010 08:11 -0700 Jeremy Chadwick
> <freebsd at jdc.parodius.com> wrote:
> 
> >I can review these statistics to see if any of the disks look like they
> >may be misbehaving.
> 
> I had to back off to 7.3-R. Unfortunately the machine is the
> 'everything' server at home (routing, dhcp, storage, mail etc.) - so
> it wasn't proving very popular messing around with it :(
> 
> I put 7.3-R back on - everything works as it did. ZFS re-scrubbed
> the pools, and I'm good to go.
> 
> We have a 'very similar' machine at the office (same controllers
> etc.) - I'll see if I can get enough drives together and run that
> up. If that fails as well it's a much better platform to debug on :)

All I'm interested in at this point are the drives in the machine which
is having the problem.  It doesn't matter if you're running smartctl on
7.3-RELEASE or 8.x -- the disk SMART stats will be the same.

So if you can provide them, I can review them and point to issues which
might explain disk I/O deadlock.

You did mention some bad sectors, and with regards to those, people very
often misread the attributes and assume the wrong thing (meaning, "oh
looks like the disk found some bad LBAs and so they're fixed" when the
situation is actually "the LBAs are bad and not fully remapped" which
can cause I/O deadlock if those blocks are read and/or sometimes written
to).

-- 
| Jeremy Chadwick                                   jdc at parodius.com |
| Parodius Networking                       http://www.parodius.com/ |
| UNIX Systems Administrator                  Mountain View, CA, USA |
| Making life hard for others since 1977.              PGP: 4BD6C0CB |