Hi,
I recently switched to kernel 2.6.32 (self configured/compiled) and I've noticed an odd behavior on one of my servers. I run a bittorrent tracker on one of my servers and when using kernel 2.6.32, every now and then ksoftirqd will jump up to 20% of my CPU for almost a minute. The tracker has a total peers amount of ~45000 peers (if it matters) and this behavior seems to only happen under kernel 2.6.32. Switching back to kernel 2.6.31.1 on the server, ksoftirdq does not behave that way.
Any thoughts?
PS: please CC me as I'm not subscribed to the list
Thanks :)
_________________________________________________________________
Windows Live: Friends get your Flickr, Yelp, and Digg updates when they e-mail you.
http://www.microsoft.com/middleeast/windows/windowslive/see-it-in-action/social-network-basics.aspx?ocid=PID23461::T:WLMTAGL:ON:WL:en-xm:SI_SB_3:092010-
On Sun, Dec 13, 2009 at 08:05:29PM +0100, Nico wrote:
>
> Hi,
>
> I recently switched to kernel 2.6.32 (self configured/compiled) and I've noticed an odd behavior on one of my servers. I run a bittorrent tracker on one of my servers and when using kernel 2.6.32, every now and then ksoftirqd will jump up to 20% of my CPU for almost a minute. The tracker has a total peers amount of ~45000 peers (if it matters) and this behavior seems to only happen under kernel 2.6.32. Switching back to kernel 2.6.31.1 on the server, ksoftirdq does not behave that way.
>
> Any thoughts?
>
> PS: please CC me as I'm not subscribed to the list
>
> Thanks :)
Hi,
I wonder if this problem still happens in 2.6.33-rc2.
A trace of softirqs during the issue would be helpful
to determine which softirq is the culprit:
mount -t debugfs /debug
echo 1 > /debug/tracing/events/irq/softirq_entry/enable
# wait a bit
echo 0 > /debug/tracing/events/irq/softirq_entry/enable
cat trace
Thanks.