DX Application Performance Management

We installed the 10.5.1.8 unix java agent last week on a couple of servers running WLS 12c; as part of the install, we modified the server start tab arguments and class path; since that time, the jvms start up anywhere from 1 to 3 minutes slower on average; our environment has 8 jvms so this is adding a considerable amount of extra time overall to our deployment process. I'm not seeing errors in the introscope agent logs. Are there any optimization tips for the java agent or can we change the configuration or introscope.profile to improve performance? 

Dear Michael:

APM 10.5.1.8 is the GA release. There are 60 or so hotfixes since then that  may help improve agent performance. 

Would need to see what is features/tracers that you have enabled.

Thanks

Hal German

We have seen that deep inheritance can delay startup so you could try to turn it off in the agent profile

introscope.autoprobe.deepinheritance.enabled=false

The feature is designed to turn off after 2 minutes so if you found turning the feature off helps, you could reduce this value (in milliseconds) to one minute (60000)

introscope.autoprobe.deepinheritance.auto.turnoff.maxtime.total=120000

There are other settings with a similar prefix in the agent profile that you could reduce the values for (for example by half)

Other things that might happen more at the beginning of the life of the application would be smart instrumentation and automatic entry point detection, so you could review these settings

introscope.agent.deep.entrypoint.enabled=false

introscope.agent.deep.instrumentation.enabled=false

These are only diagnostic suggestions, so this might lead to a support case. Having said that, if you turn off deep inheritance or can configure a suitable value for it and are seeing enough data for your needs, you can proceed like that, it is always the balance of what data you get out and what impact it may have to the monitored application.

The last note would be if you have happened to enable full instrumentation (weblogic-full.pbl) to change it to typical (weblogic-typical.pbl)

Unfortunately, also because it is an issue with startup, you would need to recycle the JVMs to test but understanding the behaviour is the same with each JVM, you could try it on separate nodes to see how it behaves.