It's all about the answers!

Ask a question

Performance Issues


Karen Steele (1.1k1126102) | asked Jan 17 '17, 8:24 a.m.

We have a single server installation of CLM using an Oracle back end database.  Throughtout the day the systems slows down to a crawl before coming back.   We have checked for anything else running on the server e.g. backups or the like there is nothing.  The installation is using the built in Liberty server engine. 

We are seeing peaks in the java entity ramping up to be using 5GB then coming back down and again we cannot determine how or why.  

There are only 15 users registered and only one clm project on the server of which only the RTC piece is being actively used.

What could be impacting this ?  Any suggestions on what we should be looking for.

As an aside, can we recreate the installation somewhere else using the same public URI and merely drop the Oracale databases on one side and restore on the other, moving all the teamserver files ?

2 answers



permanent link
Kevin Ramer (4.4k6159178) | answered Jan 17 '17, 1:39 p.m.
edited Jan 17 '17, 1:42 p.m.

In our experience, Java garbage collection can cause the application to become less responsive.   When garbage collection is running, nothing else in the JVM is allowed to run.   One could add:

-Xverbosegclog:/gcc-logs/host-name/gc#.log,100,10000

to the generic VM arguments, then use a garbage collection analyzer to look for long periods of "exclusive access".   I gather the Server Status information and I see "sawtooth" pattern of the free VM over time in the JVM


Other factors are CPU utilization ( especially if database and websphere are on the same hardware ).  Also, processes can be come I/O bound if swap space is heavily used.

It is possible to relocate the setup, but it is best if host name can be preserved.




Comments
Karen Steele commented Jan 20 '17, 3:34 p.m. | edited Jan 20 '17, 3:47 p.m.

so we uprooted the clm 6.02 installation to a new squeaky clean server - nothing on it other than windows, java and clm ... transferred the teamserver.properties and the likes and everything came back up - it was looking good until we got a major heap on java again and it crashed this afternoon.

Seeing multiple errors in the logs for java related stuff .. I know clm and installation but the java stuff and errors are beyond me ... if I took a sample would you / could you perhaps have a look ? 

Forgot to mention, there are no diagnostic errors on the server at all .. so its really baffling why this is happening. 

Config is : Windows Server, CLM 6.0.2, Liberty Websphere (built in), Liberty Basic User Registry, Oracle Databases - all clm apps except Global Configuration and RELM, single server, 20 Users


Kevin Ramer commented Jan 20 '17, 5:07 p.m.

From your descriptions I would expect Out Of Memory exceptions which should write javacore, Snap and heapdump files.    The javacore are text files and the first few dozen lines will have the exception that lead to the condition.

Depending on how much RAM your server has you may simply need to adjust JVM (java virtual machine) heap maximum.

https://jazz.net/forum/questions/222652/how-to-edit-jvm-settings-of-bundled-was-liberty-in-case-of-clm-602





permanent link
Ralph Schoon (56.0k23642) | answered Jan 19 '17, 3:08 a.m.
FORUM ADMINISTRATOR / FORUM MODERATOR / JAZZ DEVELOPER
edited Jan 19 '17, 3:24 a.m.

Which version of CLM? This is the most basic information everyone should include in their question.  How should I ask a question in the Forum if I want to receive useful answers?

My experience with 6.x versions is that some of the new components such as link indexing requires a lot of resources and it is better to distribute the applications. 

You would need some performance monitoring in order to be able to find out what the issue is.



Comments
Karen Steele commented Jan 19 '17, 10:37 a.m.

thanks for the input ... yes we have seen 6.0x being more intensive and requiring an enterprise topology versus single servers.  

Your answer


Register or to post your answer.