We are currently upgrading from v9.6 to v10.0 and we have issues with the Java agent (websphere 7 with ibm jre 1.6) with the following error in the autoprobe log
[ERROR] [IntroscopeAgent.Agent] Introscope AutoProbe will not run because the provided directives files are not valid: Can't proceed because the specified tracer "MethodCalledErrorReporter" is invalid. Please restart with valid directives.
Reviewing the required.pbd there are two trace mappings that are commented out
# SetTracerClassMapping: MethodCalledErrorReporter com.wily.introscope.agent.trace.hc2.MethodCompletedErrorReportingTracer com.wily.introscope.probebuilder.validate.MetricNameValidator
# SetTracerClassMapping: HTTPErrorCodeReporter com.wily.introscope.agent.trace.hc2.HTTPErrorCodeReporter com.wily.introscope.probebuilder.validate.ResourceNameValidatordoes anyone know why these mappings are commented out in the out of the box required.pbd for the version 10.0 java agent?Thanks,Billy
com.wily.introscope.probebuilder.validate.ResourceNameValidatordoes anyone know why these mappings are commented out in the out of the box required.pbd for the version 10.0 java agent?Thanks,Billy
I'm finding more "SetTracerClassMapping" lines being commented out from the out of the box v10 java agent
The latest is within the java2.pbd
#SetTracerClassMapping: NIOSocketSummaryBandwidthTracer com.wily.introscope.agent.trace.nio.SocketClampAwareMethodResultRateTracer com.wily.introscope.probebuilder.validate.MetricNameValidator
I suspect that there are more of the tracer mappings that are commenting out and we are uncommenting the mapping line and restarting the JVM to see what the next mapping we need to locate.
Does anyone have any knowledge on why these tracer class mapping pbd lines are commented out?
Can you let us know exactly which agent packages you are using? We've looked locally in several variants of the agent packages, and haven't yet found any evidence of the statements above so far.
In the first case (required.pbd), are these tracer lines commented out as shown by the autoprobe log or directly in required.pbd?
When in doubt, start fresh from an installer and start over. That is my lesson from this mess.
Another lesson, Good intentions with little knowledge will cost you more resources to fix.
I knew these lessons, ones that I have used quite often with APM, but I was out of the office the majority of November, with vacation and CA World 2015 (which rocked) and thought my very simple directions, tech notes and guidance was more than enough, guess not.
Come to find out one of the Jr.s on the team went full bore on trying to align our 9.6 java agent configuration with the 10.0 agent. With very bad results...well no results being metrics.
So, I pulled a full /core/config directory for all application servers, went through and copied the entire websphere-typical.pbl tree. In the toggles-typical.pbd, commented out the NIO since with WebSphere it generates hundreds of thousands of socket connections daily and in the past has caused metric explosions.
Found a test instance that I could restart the JVM without impacting too many people, deleted the pbd/pbl/profile from the <apm agent home>/core/config, copied in the new config file set, insured the EM host was set to my MOM, fired it up.
Accessed the websphere console, since the JVM was the deployment manager, and instantly saw the front end metrics.
So, if you hit a problem like this, compare what you are trying to debug with a clean copy from the installer and rule out if someone's good intentions, to take initiative, to find someone constructive to work on, is your problem.
Thanks to everyone,