Strange performance problems with I/O subsystem
Ivan Voras
ivoras at freebsd.org
Mon Mar 11 12:58:46 UTC 2013
On 08/03/2013 08:07, Гуляев Гоша wrote:
> 05.03.2013, 23:14, "Ivan Voras" <ivoras at freebsd.org>:
>
> On 27/02/2013 05:39, Гуляев Гоша wrote:
>
> And when I have some intensive I/O tasks (csup -g /root/mysupfile,
> or
> portsnap fetch update, or rm -rf /usr/obj, when system working
> sequentialy with many files in general) my system begins
> periodically
> freezes. If I see gstat it shows me 100% I\O on disk.
>
> It usually means some kind of incompatibility of your disk drive and
> the
> system, its controller or its driver.
> Sometimes it could me that the disk is broken.
>
> Ivan, thank for answer! I think there is no any hardware problem,
> because as I say before, in other OS'es all works fine, without
> performance problems. How I can diagnose problem deeply, and maybe give
> some useful information for resolving that issue?
It is hard to give good general advice in this case.
You could maybe see if there are any error messages in your
/var/log/messages file. You could also monitor the wchan of the hung
process and verify if it's stuck in the "wdrain" state (in "top"). Maybe
you could see if there is a problem with interrupts on the machine
(either by using "systat -vm" or "vmstat -i"). You could run "iostat -x"
and see if the %busy column is 100% while there are many queued requests.
If you install smartmontools, you can run a couple of tests just in case
(try smartctl /dev/ada0 and smartctl -t long /dev/ada0).
... but if the problem is an incompatibility with the driver, all this
will probably not help you much.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 260 bytes
Desc: OpenPGP digital signature
URL: <http://lists.freebsd.org/pipermail/freebsd-hardware/attachments/20130311/d5a4095b/attachment.sig>
More information about the freebsd-hardware
mailing list