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

Openshift pod killed due to instrumentation overhead on jboss-cli-client.jar

p_esposito
Contributor

Hello,
I'm having a blocking problem on Openshift due to a "unwanted" OneAgent instrumentation, and it would be very useful to exclude a specific jar file to be monitored. Can I configure OneAgent to exclude it?

Going deeper into this problem, after starting each POD, its status is "running".

When the POD is in status "running", the replication controller sends a check connect through the JBOSS CLI remote jar (jboss-cli-client.jar).

The check connect timeout is set to 10 seconds and normally, when the POD starts up without OneAgent, the response takes only 2 seconds.

When the POD is started with OneAgent, the agent injects itself into the JVM process, including the jboss-cli-client.jar which is only executed to check if POD is running.

This instrumentation takes a long time, causing a slowdown in the POD response of 30 seconds, which means it ends with a timeout after 10 seconds; this timeout lets the replica controller believe that the POD does not start correctly, causing a kill process that ends the POD.

Have you ever found this behavior? If yes, how did you solve it?

Regards

Pasquale

1 REPLY 1

p_esposito
Contributor

thanks to Roberto Vannucci, we solved the issue by excluding the jboss-cli-client.jar in settings->monitoring->monitoring overview