Memory Leak SAP PI Heap Dump

[luiscarlos_muz]

Hi everyone,

We got a Heap Dump on our PI system, using  EMA we discovered that there is indeed a leak suspect.  With this problem description:

""One instance of "java.util.concurrent.ConcurrentHashMap" loaded by "<system class loader>" occupies 1,088,886,176 (63.29%) bytes. The instance is referenced by com.wily.introscope.agent.AgentFactoryThread @ 0x70000009a81a240 Agent Heartbeat , loaded by "<system class loader>". The memory is accumulated in one instance of "java.util.concurrent.ConcurrentHashMap$Segment[]" loaded by "<system class loader>"".

It seems a Wily Introscope related problems, Does anyone else got this issue? Is there a solution we can implement?

[AntonioSanz]

Are you running SAP PI in a virtual machine? Check that, some times virtual machines have restrictions to memory, although your machine have enough memory, the virtual machine can restrict the use of it, and Java can launch memory exceptions.

Regards.

[vadimklimov]

Hello Luis,

I agree with your assumption: the issue seems to be related to a memory leak in deployed Wily Introscope bytecode agent that is used on your PI system.

Precisely speaking, the leak suspect details that you posted, look very similar to the memory leak that, for example, was identified in Wily agent version 9, SP10  and was fixed in patch 3 for it, but I'm not sure which version of Wily agent you actually use (can you share this information with us here?) and if this parch is valid for it.

Generally soeaking, there were several memory leak issues identified for various WIly agent versions during last years, so the best recommendation would be to apply latest patch to the used Wily agent version.

Regards,

Vadim

[luiscarlos_muz]

Hi Vadim,

Thanks for your reply, we are using the Introscope  release 8.2.3.5 (Build 476583).

[vadimklimov]

Hello Luis,

Can you update the used Wily agent and install the latest patch for it? For release 8, it is Wily agent version 8.2.4.0, patch 1. In patches for version 8.2.3.5, as well as in version 8.2.4.0, there were quite a lot of improvements - including fixes for some memory leaks.

If this will not help, I would suggest to create an SAP incident for component XX-PART-WILY and provide the heap dump to SAP Support for further analysis. Precise analysis of dominator tree and references for leak suspect objects should help here in identifying the root cause.

Regards,

Vadim

[Harish]

Hi Luis,

From the dump it seems one of your PI mapping is creating hash map object and occupying lot of space in the java run time memory. you need to find out the map which is using the memory (the full trace will help to identify the object.

Please refer the below sap help

Analyzing Out-of-Memory Situations - SAP Documentation

also refer the below discussion

Java Mapping - Memory Use &amp;amp; Memory Recovery | SCN

regards,

Harish