Memory leak: CMS Old Gen Heap Issue

[#1]

Hello,

Environment - AEM 6.4.8

Facing frequently CMS Old Gen Heap Issue in production publishers. By reading through the HeapDumps, it is always pointing to CacheLIRS which is taking around 65% of the space and causing issues. 

One instance of "org.apache.jackrabbit.oak.cache.CacheLIRS" loaded by "org.apache.felix.framework.BundleWiringImpl$BundleClassLoader @ 0x6c26bc280" occupies 2,666,504,552 (64.64%) bytes. The memory is accumulated in one instance of "org.apache.jackrabbit.oak.cache.CacheLIRS$Segment[]" loaded by "org.apache.felix.framework.BundleWiringImpl$BundleClassLoader @ 0x6c26bc280".

Keywords

org.apache.jackrabbit.oak.cache.CacheLIRS

org.apache.felix.framework.BundleWiringImpl$BundleClassLoader @ 0x6c26bc280

org.apache.jackrabbit.oak.cache.CacheLIRS$Segment[]

HeapDumps & thread-dumps are pointing towards the below thread. 

•    <Java Local> java.lang.Thread @ 0x6ea66f4e0 HealthCheck Sling/Granite Content Access Check Thread

Do we require the HealthChecks monitoring dashboard in Production Publishers? there are other agents like NewRelic are already installed. 

[#2]

org.apache.jackrabbit.oak.cache.CacheLIRS - is not an issue at all.

I suggest opening a support case and share thread & heap dump along with log files

[#3]

Thanks for your reply. Adobe Support Case has already been created for this issue, I thought of checking some of the experts opinion in this thread if there are any other pointers in terms of fixing the issue.

[#4]

If you aren't, please ensure AEM is running with at least these 4 JVM flags : 

-XX:+PrintGCApplicationStoppedTime -XX:+PrintGCDateStamps -XX:+PrintGCDetails -XX:+PrintGCTimeStamps

Which will have Java write some GC information to disk :

gc.log, gc.log.1, gc.log.8.current etc

Provide those logs if you have them to the case you raised.

[#5]

If you aren't, please ensure AEM is running with at least these 4 JVM flags : 

-XX:+PrintGCApplicationStoppedTime -XX:+PrintGCDateStamps -XX:+PrintGCDetails -XX:+PrintGCTimeStamps

Which will have Java write some GC information to disk :

gc.log, gc.log.1, gc.log.8.current etc

Provide those logs if you have them to the case you raised.

[#6]

Thanks for your reply. Yes, those flags are present, since it is AMS platform, I do not have access to those logs. CSE has already provided those logs in the Adobe Case. I hope, this issue will be resolved soon.

[#7]

Hi,

I am not totally sure if that's a normal behavior. Can you check in /system/console/jmx if the number of "SessionStatistics" entries is growing over time. In that case you have a session leak, which often goes together with memory issues like you described.

[#8]

Hello @Jörg_Hoh,

Could you please explain what do you mean by "he number of "SessionStatistics" entries is growing over time"? Now I've checked our number and it is 104. If it will increase in some hours, can it be session leak?

Thanks.

[#9]

104 is fine.

This generally is not of concern until the jcr-session count goes into the thousands.

[#10]

Did you get any solution for this?