Formula of how OA agent calculates CPU, Memory, Swap utilization
OA agent 12.04 on all unix flavours
We want to know how OA agent calculates per CPU, Global CPU, Memory and Swap utilization.
Please help us since end-users are complaining that CPU, Memory and Swap utilization is not accurate.
End-users have their own formulas to compute CPU, memory etc and they use this formula and when they find an alarm from OMi policy differs, they are raising.
Hence we want to know how OA agent collects performance metrics for CPU etc.
The information about performance metrics is documented here:
If you want more details contact the support team.
my customer users complain because top and vmstat are realtime tools. so this is what can recommend
1) OA Agent collects performance data every 300s. use the parm file to increase intervall of collected data to 15s . There is a policy in Infrastructure MP so you can deploy the settings on all agents.
2) this document helps to understand why top and OA agent data collectons differs:
3) use the metric dictionary of the OA Agent to understand what the metrics mean. i build dashboards with all metrics and the definitions side by side. so left the graphs , right the definiton. so its easier to identify which metrics are useful.
4) build your own dashboards as HP default Dashboards are not user friendly. e.g. change datapoints to raw so you can make use of the data collection intervall of 15s
5) if you use omi enable the realtime license of the agent (oalicense tool) so you can use the realtim emetrics. this license is included in opsbridge premium.
the features of the performance dashboards in Omi is very good, unfortunately most is not documented at all, especially how to build great dashboards. so reverse engeneering and try and error is the only option. not to mention that i found some defects on my journey. i build dozens of dashbord for one of my custumer, a global car manufacturer. now they start to get happy and ready to use it in daily operations.
thats easy to explain.
the data collection intervall is at OA Agent 300s while User use top or vmstat.
the default Performance Dashboards are not really helpful as these lacks metric descriptions.
and understanding OA performance metrics and creating dashboards is painful as documentation is incomplete, misleading, and some stuff isnt just documented at all.
also i found some metrics collected by the agent are definitly inaccurate. HPE support is working on this.