MachineAgent does not detect JVM Crash

[dhyunkang]

HI all,

I tried to detect JVM Crash events.

First, I configured Policy for JVM Crash as mentioned in documentation like the attachment.

And I started MachineAgent.

I killed JVM with -9 option. Nothing was reorted in Events screen.

When I restarted the JVM, JVM restarted event was displayed.

And when JVM is killed, this message was logged in machine-agent.log file.

[Agent-Monitor-Scheduler-3] 27 Mar 2015 11:34:14,919  INFO SigarAppAgentMonitor - Detected that JVM for PID 15572 is no longer active

My platform is CentOS 5.5 64 bit.

I use JDK 1.6 32 bit version for both AppServerAgent and MachineAgent.

For controller, I use JDK 1.7.0_55 64 bit version.

Can anybody help me to fix this?

Regards,

Scott Kang

[Arun.Dasetty]

Hi Scott,

Can you check if information in doc link and section listed below from doc link , is the case here?

https://docs.appdynamics.com/display/PRO40/JVM+Crash+Guard : "If an uninstrumented JVM crash happens within less than a minute of a previous crash then it will not be reported by the Standalone Machine Agent. In some circumstances, the JVM may crash and then be restarted only to crash again within one minute. For this repetitive cycle crash and restart scenario, only the first JVM crash is reported by the agent."

Please send the machine agent logs if you feel otherwise

Regards.

Arun

[dhyunkang]

Thank you for your answer.

The JVM I tested is already in the scope of monitoring. The JVM is in the chain of business transactions. 

I wanted to test how I can get  notifcation from the crash so that I can execute some script.

I think it may be related with some system environments like OS version, system libraries, and etc.

I could not get enough clue from the log of machine agent.

How can I investigate further?

Thanks again.

[Arun.Dasetty]

Hi ,

Can you confirm whether the jvm crash case is falling under case description that i listed in previous update Or confirm whether this the case "if an uninstrumented JVM crash happens within less than a minute of a previous crash then it will not be reported by the Standalone Machine Agent"

If not please send the archive version of machie agent logs folder for our initial review ?

Regards,

Arun

[hitesh_shetty2]

Im facing the same problems, I've configured the health rules to detect a JVM crash with policies and actions in place. Only there are no emails sent (per action) when I kill the JVM with a -9 option. Any thoughts ?