[CentOS] Tracking down whats causing a high load?

[Peter Kjellström]

On Wednesday 21 June 2006 12:17, Ian mu wrote:
> Hiya,
>
> Currently running Centos 4.2 x86_64 dist on a dual 3G xeon, 2G ram,
> scsi setupand everythings been running fine on it for some time. Then at
> 4am last night something kicked in (have mrtg running monitoring when) and
> since then its been running a load of about 1.5 (normally around 0.4). CPU
> usage is Cpu(s):  1.1% us,  0.6% sy,  0.0% ni, 97.9% id,  0.2% wa,  0.1%
> hi,  0.1%si.
>
> Can't see any new processes that would cause the load, just wondering is
> there any way to try and track down whats actually causing this? It's not
> excessive load, but want to add some new services and wary now, its
> something that seems wrong given the sudden increase at 4am (think thats
> when some o.s housekeeping tasks are normally scheduled, but there's none
> running that I can see that started today).
>
> Just hoping someone may have some tips on checking whats always waiting or
> how to isolate whats happening.

hit it with the big sledgehammer, oprofile :-) that will very likely tell you 
if the kernel is doing something (besides the expected mwait_idle-ish...). 
rough guide (not for cut-n-paste):

yum install oprofile kernel-smp-devel (and manual install of kernel-debuginfo)
opcontrol --setup --vmlinux=...vmlinux from -debuginfo
opcontrol --reset ; --start ; sleep ; --stop
opreport -l -p /lib/modules/$uname -r | head

/Peter

> As said, ps -ef shows no new processes, and 
> cpu usage is very low.
>
> Tia, Ian
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 191 bytes
Desc: not available
URL: <http://lists.centos.org/pipermail/centos/attachments/20060621/ad1ae2d9/attachment.sig>