[CMIS-868] OOM issue when run concurrent test on GetDescendant, GetTypeFlat, GetTypeTree.

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: OpenCMIS 0.12.0
• Fix Version/s: OpenCMIS 0.13.0
• Component/s: opencmis-client

Environment

AIX 7.1+Websphere ND 8.5.5.3
Java(TM) SE Runtime Environment (build pap6460_26sr2ifix-20120419_02(SR2+IV19661))
IBM J9 VM (build 2.6, JRE 1.6.0 AIX ppc64-64 20120322_106210 (JIT enabled, AOT enabled)
J9VM - R26_Java626_SR2_20120322_1722_B106210
JIT - r11_20120322_22976
GC - R26_Java626_SR2_20120322_1722_B106210
J9CL - 20120322_106210)
JCL - 20120316_01

• Assignee: Florian Müller
• Reporter: Yi Wang
• Created: 14/Nov/14 12:07
• Updated: 20/Nov/14 16:46
• Resolved: 20/Nov/14 16:43

Description

When I run concurrent test with 100 threads on GetDescendants, GetTypeTree, GetTypeFlat function.

I found there are many hashtable objects(>500MB) saved in heap after analyzed heap dump. I found there was one line(red line below) in convertTypeDefinition method that chemistry used typeDefinition object as the key of HashMap. However, JAVA does not think 2 typeDefinition objects are same objects although the value of typeDefinition members are same. I debug to this line, the "result" variable always was "null". It made the objectTypecache larger and larger when run concurrent test.

private ObjectType convertTypeDefinition(TypeDefinition typeDefinition) {
lock.writeLock().lock();
try {
ObjectType result = null;
if (objectTypeCache == null)

{ objectTypeCache = new IdentityHashMap<TypeDefinition, ObjectType>(); }

else

{ result = objectTypeCache.get(typeDefinition); }

if (result == null)

{ result = objectFactory.convertTypeDefinition(typeDefinition); objectTypeCache.put(typeDefinition, result); }

return result;
} finally

{ lock.writeLock().unlock(); }

}