Memory leak:Getting heap dump at my AEM publisher

[swapnilu7465684]

Hi Team,

I am getting heap dump and GC on my AEM server when doing soak test the following is the report.

Problem Suspect 1

One instance of "org.apache.sling.models.impl.ModelAdapterFactory" loaded by "org.apache.felix.framework.BundleWiringImpl$BundleClassLoaderJava5 @ 0x5c7e3e280" occupies 214,336,312 (29.68%) bytes. The memory is accumulated in one instance of "java.util.concurrent.ConcurrentHashMap$Node[]" loaded by "<system class loader>".

Keywords
java.util.concurrent.ConcurrentHashMap$Node[]
org.apache.felix.framework.BundleWiringImpl$BundleClassLoaderJava5 @ 0x5c7e3e280
org.apache.sling.models.impl.ModelAdapterFactory

Details »

Problem Suspect 2

10,992 instances of "org.apache.jackrabbit.oak.segment.Segment", loaded by "org.apache.felix.framework.BundleWiringImpl$BundleClassLoaderJava5 @ 0x5c1c84e10" occupy 79,999,744 (11.08%) bytes. These instances are referenced from one instance of "com.google.common.cache.LocalCache$Segment[]", loaded by "org.apache.felix.framework.BundleWiringImpl$BundleClassLoaderJava5 @ 0x5c1c77e70"

Keywords
org.apache.felix.framework.BundleWiringImpl$BundleClassLoaderJava5 @ 0x5c1c77e70
com.google.common.cache.LocalCache$Segment[]
org.apache.felix.framework.BundleWiringImpl$BundleClassLoaderJava5 @ 0x5c1c84e10
org.apache.jackrabbit.oak.segment.Segment

Can someone tell me how to proceeds

[wimsymons]

According to what I see, is that you only have 800mb of heap space. Seems to me you did not pass Xms and Xmx parameters to your publish instance. A production instance needs more than 1gb ram. Give it some more breathing room. For example 4gb.

[swapnilu7465684]

Hi Thanks for your response

we have started JVM as per below

java -server -Xms8192m -Xmx8192m -XX:MaxPermSize=256M -Djava.net.preferIPv4Stack=true -Djava.awt.headless=true -XX:+UnlockDiagnosticVMOptions -XX:+HeapDumpOnOutOfMemoryError -XX:+PrintGCDetails -XX:+PrintGCDateStamps -XX:+PrintTenuringDistribution -Xloggc:/mountpoint2/aem/crx-quickstart/logs/gc-2019_01_05-09_08.log -Dsling.run.modes=publish,uat,standalone,nosamplecontent,crx3,crx3tar -jar crx-quickstart/app/cq-quickstart-6.3.0-standalone-quickstart.jar start -c crx-quickstart -i launchpad -p 61003 -Dsling.properties=conf/sling.properties

[swapnilu7465684]

Kindly tell me if anything need to change in above start up script.

[wimsymons]

Strange. Looks legit to me.

Can you take a look at http://<youraemhost>:4502/system/console/memoryusage (the URL, not the path) and paste the content?

That should show your current memory configuration and free space.

[swapnilu7465684]

Please see

[wimsymons]

Looks totally fine. (We have similar settings).

But your suspects are strange as they consume so little heap space. (214 MB for suspect 1, 79 MB for suspect 2).

Are you sure you didn't miss something else?

When we have OOM's on AEM, we find suspects holding 6 GB or more.

[swapnilu7465684]

That is why I have to see whether it is a really need to worry and look the solution for it.  The full GC report are as per below

Class Name	Shallow Heap	Retained Heap
org.apache.sling.models.impl.ModelAdapterFactory   @ 0x5c7ee6ac8	96	214336312
org.apache.felix.framework.BundleWiringImpl$BundleClassLoaderJava5   @ 0x5c948db38	96	47600992
org.apache.jackrabbit.oak.plugins.index.lucene.IndexTracker   @ 0x5c24d1dc8	40	36948344
org.apache.jackrabbit.oak.cache.CacheLIRS   @ 0x5c25478a0	64	32506632
org.apache.jackrabbit.oak.segment.file.FileStore   @ 0x5c1de28c8	112	29710512
org.apache.jackrabbit.oak.cache.CacheLIRS   @ 0x5c1e23f30	64	21838752
org.apache.felix.framework.BundleWiringImpl$BundleClassLoaderJava5   @ 0x5c2484e60	96	12621072
com.adobe.granite.ui.clientlibs.impl.HtmlLibraryManagerImpl   @ 0x5c37cc4d0	64	10548568
org.apache.sling.scripting.sightly.js.impl.jsapi.SlyBindingsValuesProvider   @ 0x5c9cb2378	48	10136152
org.apache.jackrabbit.oak.segment.SegmentTracker   @ 0x5c1de2938	24	7718544

[wasil_z]

There have been couple of improvements included in AEM 6.3 with regards to ModelAdapterFactory(6.3.2.1)[2], so please try on latest AEM 6.3 SP + CFP, taking into account the latest Apache Sling Models Implementation[1]

Thanks,

Wasil

[1] Maven Repository: org.apache.sling » org.apache.sling.models.impl

[2] Adobe Experience Manager Help | Release Notes: AEM 6.3 Cumulative Fix Pack

The key highlights of the AEM Cumulative Fix Pack are:

• The built-in repository (Apache Jackrabbit Oak) is updated to version 1.6.11.
• Improved rendering of Parsys components in Classic UI.
• Updated sling models bundles to include character set fix.
• Made publish to Brand Portal async for all assets types.
• Updated coralui-component-richtexteditor.git from 0.1.15 to 0.1.16
• Fixes in show/hide functionality of drop-down component.
• Enabled image flipping for the Core Image Component.
• Updated felix http bundles to enable session attributes.
• Removed cache=true on Sling Models due to memory consumption issues.