From: ldg@ulysium.net
Date: Thu Jun 12 2003 - 22:37:57 EDT
In my original post (see snippet below), I was asking how to track down what
thread(s) would be consuming cpu time in the kernel while it's idling.
I tried many things and got many different suggestions, such as various ps
commands to get more detailed info about the threads running, but I couldn't
find what those threads belonged to because those different commands don't
give enough info in this case to find out for sure what is causing this.
Basically, the system idles most of the time, as it isn't host of too many
things yet, but the cpu usage from the idling kernel always shows 10% to
20%, and the load average between 0.1 and 0.2, all this while almost no
activity is registering in any other processes.
My only possible suspect was vfast, which I run on 5 domains, including root
and usr, but I couldn't get any indication that vfast actually was causing
this.
I was then shown how to use dbx to track this down better in the kernel and
found a relation to vfast in the cpu consuming thread(s).
So I tried suspending operation of vfast on the root domain, which didn't
change anything, but then suspended vfast on the usr domain, and sure
enough, the cpu usage gradually dropped back to near zero, as well as the
load average.
So it is confirmed that vfast is causing this cpu usage and slight load even
during extended periods of system inactivity.
The one thing that still doesn't make sense to me, is that with a lot of
very long periods of system inactivity, vfast runs all the time and never
gets finished doing its job.
Although I limited the IO usage of vfast to 1% during busy operation, this
doesn't apply here, as the system is actually idling most of the time.
I'm inclined to wonder how vfast will affect overall system performance
during high system load, if it can't even finish its job in a reasonable
amount of time while the system is completely idle, then how can it ever
expect to keep things tidy during higher loads???
>> USER PID %CPU %MEM VSZ RSS TTY S STARTED TIME
>> COMMAND
>> root 0 12.5 2.9 1.37G 37M ?? R < 11:58:09 49:16.12
>> [kernel idle]
>>
>>
>> load average: 0.14, 0.10, 0.10
>>
>> The load average stay a lot around 0.10 and 0.20 even when most tasks are
>> shut down and there's no load on the system.
-- Didier Godefroy mailto:dg@ulysium.net
This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:49:22 EDT