cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

Any idea about Dynatrace Instrumentation CPU overhead issues?

mallempati_sree
Inactive

hi All,

While running Stress test we saw servers reaching 100% cpu. And when doing CPU sampling it shows dynatrace instrumentation CPU overhead is high and thread dump show that it has 29 blocked threads by Dynatrace.diagnostics.agent.af method. Any idea about these issues?

Thanks,

Sree

4 REPLIES 4

Joe_Hoffman
Dynatrace Champion
Dynatrace Champion

This can likely be caused by a misconfiguration. I would suggest removing any custom instrumentation, setting JDBC to Aggregation, logging sensor to aggregation, Exception sensor to not capture stack, ensuring String capture length is small (256?), then retesting. If it still happens, remove sensors for JDBC, Logging, Exception. The goal is to determine which sensor is causing the problem. You might be able to get some hints about which sensor is involved from the thread dump you mentioned, look at the stack above the dynatrace agent. JDBC? Once you determine the offending sensor or setting, you can adjust accordingly.

Thanks Joseph!

I will check these. It's much helpful info than the one suggested by support for switching Agent platforms rather than finding Root Cause.

Thanks,

Sree

HI Joseph,

Is it possible to create a dashboard with any measure for all the agents in the system to identify dynatrace CPU instrumentation Overhead in Appmon v7.0?

last time we identified it when we did sampling for few servers. as we are doing stress test CPU will be 100% and would like to see if any of the agents has any overhead caused by Appmon.

Thanks,

Sree

Joe_Hoffman
Dynatrace Champion
Dynatrace Champion

Sree, It is very hard to quantify the CPU overhead of AppMon. The best way is to run your system WITH AppMon, then run it without AppMon. But as stated above, if you see any significant CPU growth when AppMon agent is enabled, then I would first look at misconfiguration problems.