DX Application Performance Management

Hi All,  We are brand new Wily customers, heck brand new to APM as a whole, and our new WebSphere environment utilizes the new Virtual Enterprise (VE).   We have seen the Introscope agent causing some memory leaks which eventually crash the applications in question.   Is anyone else running Introscope 8.1 agents on WebSphere VE instances?   If so did you experience any problems such as our famous memory leak?   Any help would be greatly appreciated.  Thanks,  -Me

Hi,  Below is info from Support that hopefully helps!  Regards,Chris  
Running Wily agent uses
additional heap on the JVM (Java Virtual Machine) i.e. not cost free. The
additional heap really depends on what the application does,  the level of
Wily instrumentation turned on, number of Wily add-ons installed, etc.
 For example, SQL monitoring is turned on by default and if the
application is SQL intensive, it will use more heap than an application that
does very little SQL. If the additional heap  + the heap required by the
application  > the max heap specified for the JVM, this can cause
OutOfMemory error or instability in the JVM. However, this is not a memory
leak; the heap required to run Wily agent can be tuned to something very low if
needed by turning off unnecessary instrumentation.

 

A memory leak is when the heap
increases over time due to unused objects in the JVM not being released, thus
causing these unnecessary objects to pile up, and the JVM to use more memory
over time. No tuning will help with this situation. We’ve not aware there’s
such issue with 8.1 agents.

 

Best to open an incident with support so it can be investigated if it’s
a higher than expected memory usage, or a memory leak issue.    

Agreed.     We havent seen any memory leaks with Introscope - but some apps suffer from metric leaks.   Check the metric counts for your agent and see if it is increasing.   You might need to turn off JSPs/Struts or SQL metrics to reign it in.

Thank you to all who have responded to this thread.   The memory leak turned out to be present in our distributed websphere world as well.   The leaky memory was caused by an incorrect JVM agrument.   We also encountered a different problem in our WebSphere VE instances which (although we haven't applied the fix yet) crashed the apps a good while after the Wily JVM args were put in place.   This problem was alledgedly caused by an incorrect reference to the LoadRunner Field Pack file in the j2ee.pbd. It was our mistake it turns out.   Thanks again for everyone's input.

I am having similar issue w/ increase JVM Heap usage after I've installed CA Wily Introscope agent for WAS. So far the only thing we have done is clamp the metrics to 5K. But is there a solution where it doesn't use that much heap w/o sacrificing metrics and reducing instrumentation gathering?

Also, the JVM Heap jumped 3x after I installed the agent.

Thanks

Manish