Re: [gentoo-user] High load, idle CPU?

[Joshua Murphy]

On Fri, Dec 18, 2009 at 2:38 PM, Florian Philipp
<lists@f_philipp.fastmail.net> wrote:
> Hi list!
>
> My virtual server seems to have a problem and I don't know how to find it.
>
> Today, while trying to create a new postgresql database I noticed that
> certain system operations seem to take ages. Normal work on the shell
> works just fine but for example accessing the postgresql server through
> psql, restarting tomcat, postgresql or zope all show the very same behavior:
> Execution freezes, the load average climbs to 1-1.9 and after a few
> minutes (maybe longer) the task finishes correctly. During all this
> time, top reports that the CPUs are 100% idle.
>
> Other services, for example rsyncd or openvpn, work normally. I also
> tried restarting the VM. Shutdown took maybe 10 or 15 minutes. Booting
> was much faster with only a few minutes until all services were up and
> running. However, it didn't solve the problem.
>
> So, my question is: How can I find out what is causing this? Can I
> somehow trace it? Find out why a process waits?
>
> Keep in mind that this is a virtual server. I have no control over the
> kernel. Oh, and before you ask: The only recent update was rkhunter from
> 1.2.9-r1 to 1.3.4-r2. I also enabled the chkrootkit weekly cronjob.
>
> Thanks in advance!
> Florian Philipp

Well... it'll cause a MASS of output to sift through .. but strace is
likely the most powerful tool for the job of simply seeing what a
process is calling and when (notably, what call is left waiting to
return as the whole process blocks). I'd also take a look at iotop,
though the fact that top shows the cpu as idle rather than sitting in
iowait makes me doubt that iotop would tell you too much.

-- 
Poison [BLX]
Joshua M. Murphy