Monitoring Linux Run Queue on HPOA v8.6.xx
I cannot for the life of me find out how this is done. There is no policy in the OSSPI, so is it done?
Hi, you need to write a policy to monitor the metric GBL_RUN_QUEUE. Usually the warning threshold for this on Linux is the number of cores/processors + 1. i.e. if there are 4 cores/kernel threads available then the lowest threshold is 5. I would wait to check if this lasts for at least 3 intervals (15 mins, say) to send out an alert. Set higher thresholds at 6, 7,...
Rather than look at the GBL_RUN_QUEUE which is a value that is obtained just at that point, you can also look at the GBL_LOADAVG or GBL_LOADAVG5 metrics which give the state of the system summarized over a given period of time. I would keep the same thresholds in this case.
Finally, you could also use the couple of policies with the InfraSPI (specifically SISPI) which monitor GBL_LOADAVG metric in a thresholded way or using adaptive baseline approach.
If you plan to start using the InfraSPI i would recommend looking at feasibility to use the latest 11.1x agents - the InfraSPI is bundled in with this agent package.
What is the threshold value? Client seems to think it can be set to 50.....? I think they have missed how Run Q is measured, do you know or can you explain it to me please?