Clint Pachl wrote:
>But Philippe is noticing this behavior even in single user mode, right? In
>single user, init and a shell should be all that is running in userland.
Right. Even after cold-booting straight into single user mode I still
see those weird-looking load peaks.
>If in single user, I would suspect hardware interrupting the kernel. Make
>sure your monitoring tool isn't the culprit.
In single user mode I just used: while true; do uptime; sleep 10; done
Anyway, last night I cold-booted the machine into single user mode and
let it do nothing for 8.5 hours (not even the while loop above) and
then "top -S" showed the following result:
PID USERNAME PRI NICE SIZE RES STATE WAIT TIME CPU COMMAND
88820 root -22 0 0K 21M sleep - 506:34 0.00% idle0
23790 root 10 0 0K 21M sleep acpi0 0:05 0.00% acpi0
37552 root 10 0 0K 21M sleep bored 0:05 0.00% systq
15761 root 10 0 0K 21M sleep bored 0:01 0.00% systqmp
48025 root 10 0 0K 21M sleep pftm 0:01 0.00% pfpurge
0 root -18 0 0K 21M sleep schedul 0:01 0.00% swapper
68258 root 10 0 0K 21M idle usbtsk 0:01 0.00% usbtask
1 root 10 0 316K 160K idle wait 0:01 0.00% init
18103 root -18 0 0K 21M idle kmalloc 0:01 0.00% kmthread
[...]
So nothing out of the ordinary. Conclusion: this is probably just yet
another case of the load not actually representing the actual CPU
usage. Some as yet unknown kernel thread probably wakes up every
minute or so, goes into the run queue, is counted towards the load,
and then does nothing at all with the CPU... It does make the load
look weird though. Anyway, case closed; sorry for the noise.
Cheers,
Philippe
