I am trying to work out a way to monitor the Kahuna domain on our filers to identify when we have Kahuna cpu bottlenecks.
We have Performance Advisor running and have data going back until the start of the year for trending.
I know that the total Kahuna domain is Kahuna + Wafl_Ex(Kahu), am I right to assume that processor:domain_busy:wafl_exempt running on the same processor as processor:domain_busy:kahuna represents the Wafl_Ex(Kahu) process?
Based on this assumption, I see the following options:
1 - Monitor the aggregation of processor:domain_busy:wafl_exempt and processor:domain_busy:kahuna on the kahuna processor. (Assumes that the kahuna processor remains constant)
2 - Monitor the minimum value of processor:domain_busy:idle for the kahuna processor. (Assumes that all other activity on the processor is kahuna)
Has anyone any comments/tips on this approach to monitoring the total Kahuna domain? Or other ways to do it, we are unable to script it because "sysstat -M" requires elevation to diag mode..