01-10-2011 06:16 AM
Hi there,
some few days ago we ran into a full /opt-filesystem on out solaris servers
running LMS 3.2.
The source of the eating-up of the filesystem was daily creation of /opt/CSCOpx/java_pidxxx.hprof (Java Heap Dump) Files
at about 04:30AM.
Each file has a size of about 500-600MByte.
Is that normal behavior of LMS system.
Does anybody of you share the same experiences?
Any idea what is going on?
Thanks for any link
Lothar
01-14-2011 07:22 AM
Most of the Java processes in LMS are configured to dump the HotSpot profiling data on a memory exception. If you are experiencing application problems, it may be worth analyzing some of these files to see if there is a memory or scaling problem. Else, the files can simply be deleted. Doing so will not break anything in LMS.
01-16-2011 10:57 PM
Hi Joe,
thanks for the update.
Do you have any suggestions how to analyse the Heap Dumps, as files-size per dump is about 500MB?
Thanks a lot for your feedback.
Lothar
01-16-2011 11:04 PM
These files are analyzed using tools like jmap. See http://www.oracle.com/technetwork/java/javase/memleaks-137499.html#gbywi for some details. This is not something we typically recommend customers do, though. Reproducible problems should be brought to TAC for development analysis.
01-16-2011 11:44 PM
Hi Lother,
most properly you have created a job that need more memory allocation for this job. So, take the number of this job from the file name itself. then search on your jobs about it. to make sure it's the issue, stop this job. then we can see how could we increase the heap size
BR
Wael
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide