kcmond process

[Marco Di Ianni]

Hì all,

I've a RAC cluster on HP-UX 11iv2 (11.23) and during stress-test on oracle 10g I've notice that, on both nodes, the process "kcmond" (for Kernel tunable parameters monitoring) bring a cpu (of 8) to 100% utilization.
I tried to restart the daemon and the server too, but the problem persist.
I searched for a patch, but I cannot find it.

Can you help me?
Thanks!

[Victor BERRIDGE]

Hi Marco,

I may be completely wrong for I have no 11.23 to check and see... but on 11.11 kcmond is linked with ems stuff...
So I would go and stop /start the clean way EMS. That is stoping first the EMS monitoring, then the diagnostics killing afterwards the remaining EMS processes etc... and restarting the lot again


All the best
Victor

[Marco Di Ianni]

Thank you Victor,
I tried to stop EMS agent and diagmond, but the problem persists.

The process kcmond go up and down in the scale of process on top, from about 1-2% to 100% of a cpu utilization, independent of EMS and diag.

Others ideas?

[Carlos Roberto Schimidt]

Hi Marco,

Try run kcalarm and see you have alarm configured.

If you have, try delete or desactive.

Schimidt

[Marco Di Ianni]

Thank you, Carlos!!

With command "kcalarm -l" I've check that there aren't alarms for Kernel Parameters...

When the process kcmond is near to 100% CPU util. the kcalarm command doesn't work ..and reply with this output:

#kcalarm -m status
ERROR: Communication with EMS failed.

Instead, when kcmond get down in top, the kcalarm work fine.
Anyway, after "kcalarm -m off", the process kcmond keep running but with very small CPU utilization.

Now, the question is: Why, when the kernel monitoring is on, the the daemon use so much CPU?

Thanks!

[Carlos Roberto Schimidt]

I don´t have sure, but try use kcusage command to identify wich kernel parameter have utilization percent high.
Maybe there are bottleneck in any parameter, and monitoring this cause high cpu utilization.