slow hard drives crushing interactivity

Bill Rugolsky Jr. brugolsky at telemetry-investments.com
Tue Jan 25 22:01:05 UTC 2005


On Mon, Jan 24, 2005 at 06:01:55PM -0500, Havoc Pennington wrote:
> On my IBM X31 laptop, the system entirely locks up when there's a lot of
> disk access, some common situations are:
>  - when getting heavily into swap due to a runaway process
>  - when running rpm/yum
> 
> It's not *technically* locked up (i.e. if you wait long enough it will
> come back) but in practice you have to reboot if a process has a memory
> leak, and you can't do any work while running yum.

What kernel are you running?

Over the past two months I've put together our 2.6.10-based production
kernel from lots of VM, scheduler, and latency patches cherry-picked
from -mm and lkml; most of them (especially the VM fixes) are now in
2.6.11-rc2.  I do some testing on my Fujitsu laptop (Pentium M 1.6GHz,
768M RAM, 4500 RPM disk), and 2.6.10-mm*-derived kernels respond much
better than vanilla 2.6.10 under load, or when doing a massive set of
package updates.

So trying the Rawhide kernel might be an option, though the 2.6.11-bk
tree has had quite a lot breakage of late, including memory leaks. :-/
[I haven't been tracking Fedora/Rawhide kernels since around 2.6.7-1.492,
 so I don't know how things are working there.]

If you feel like doing some patching, you might also want to have a look at
Ingo's swapspace-layout-improvements patch, which AKPM added to 2.6.11-rc1-mm2,
Andrea's new OOM-killer patches [since long delays killing a process make the
OOM killer go a bit berserk],

http://groups-beta.google.com/group/linux.kernel/browse_frm/thread/9091db6e91a07f0c

and Jens's CFQ time-slice elevator,

http://www.kernel.org/pub/linux/kernel/people/axboe/patches/v2.6/2.6.11-rc1/cfq-time-slices-20.bz2

Regards,

	Bill Rugolsky




More information about the fedora-devel-list mailing list