[
https://issues.jboss.org/browse/ISPN-868?page=com.atlassian.jira.plugin.s...
]
Galder Zamarreño commented on ISPN-868:
---------------------------------------
Forget about EhCache, there appears to be a leak in Infinispan and there's a fairly
good chance that the leak was recently discovered in ISPN-845:
I took a heap dump once passed the 3500 object mark and I see this:
"3,530 instances of
"com.arjuna.ats.internal.jta.transaction.arjunacore.AtomicAction", loaded by
"sun.misc.Launcher$AppClassLoader @ 0x10a802050" occupy 235,563,912 (89.75%)
bytes. These instances are referenced from one instance of
"java.util.concurrent.ConcurrentHashMap$Segment[]", loaded by "<system
class loader>"
These ConcurrentHashMap belongs to
org.infinispan.transaction.xa.TransactionTable.localTransactions map and ISPN-845
precisely points to a leak there. I'm attaching a screenshot of Eclipse MAT.
Running out of memory using Infinispan after adding a small number of
entities
------------------------------------------------------------------------------
Key: ISPN-868
URL:
https://issues.jboss.org/browse/ISPN-868
Project: Infinispan
Issue Type: Bug
Affects Versions: 4.2.0.Final
Environment: JBossJTA 4.14.0/Hibernate 3.6.0.Final/Spring 3.0.5
Reporter: Tom Waterhouse
Assignee: Galder Zamarreño
Fix For: 4.2.1.Final
Attachments: ISPN-868-1.zip, ISPN-868.zip, ispn686-error-galder.txt,
jconsole-mbeans.JPG, jconsole-memory-view.JPG, test-progress.JPG
While running a load test data builder for our application we ran out of memory very
quickly. A simple test case (attached) was created to duplicate the issue. We found
running the simple test case illustrates that Infinispan uses a large amount of heap
space.
As a reference the same test was run using EHCache 2.2. Memory usage was much lower; we
never ran out of heap space. Note that EHCache was used as a reference only, our goal is
to go to production with Infinispan.
--
This message is automatically generated by JIRA.
-
For more information on JIRA, see:
http://www.atlassian.com/software/jira