I need advice on the frequency of the ‘Date Modified’ for AdaptiveProcessingServer i.e. how often should it change.
Today I have noticed it has changed twice within an hour. Is this normal and down to user activity?
The reason why I ask is, JAVA_PID****.hprof files started getting created in the Logging folder and filled it up. So I switched off the creation of these files.
What I did was change the command line for the AdaptiveProcessingServer:
XX:+HeapDumpOnOutOfMemoryError
to
XX:-HeapDumpOnOutOfMemoryError
Recently there has been changes to our enviroment, implemented by the environment team (not me), and the JAVA_PID****.hprof files went mental. So I took the above steps to try and sort the issue out.
Have I followed the correct actions. I swimming in waters unknown at the moment.
Also, the SIA was bounced last night and scheduled for a weekly bounce.
The APS can reboot itself when the memory reaches certain thresholds. The theory is that since Business Objects is now a 64 bit application, that it should be able to manage the memory itself.
One of my clients has been experiencing the production of the extremely large hprof files. There are a few SAP Support documents about this topic.
Do you mind me asking what specific changes were made as per your post below?
Right click on AdaptiveProcessingServer, select properties
NOTE: Back up the ‘Command Line Parameters’ box text before you do ANYTHING!
In the ‘Command Line Parameters’ box, copy the text, paste into notepad and look for -XX:+HeapDumpOnOutOfMemoryError
If you change the plus(+HeapDumpOnOutOfMemoryError) sign to a minus(-HeapDumpOnOutOfMemoryError) this will stop the JAVA_PID*****.hprof files getting written out.
Copy the amendment and paste the code back into the ‘Command Line Parameters’ box.
Also, in the ‘Command Line Parameters’ box you will be able to see the directory they get written out too and possibily change it to another destination. I have not tested that so wont say it works but its an idea.
Check all the servers for -XX:+HeapDumpOnOutOfMemoryError and repeat the above steps, if found. I found it lurking in the WebApplicationContainerServer.
Bounce affected servers since they will go into a ‘Stale’ status.
I did this on Monday and I have not had a JAVA_PID*****.hprof created.
NOTE:
This stopped the files I cannot say why they started since the system was stable for over a year.
This seems the most likely culprit. Now that you mentioned it, there are maximum memory settings for each of the Adaptive Processing Servers (assuming you have split them out per recommendations). If you get a heap dump, you may want to consider increasing the maximum memory setting for the related Adaptive Processing Server and see if that corrects it. Also, you need to look at the overall Java Heap size setting for Tomcat.
We have some heap files being created sporadically ourselves. I’m going to have to look into this also.
Well, I followed a very straight forward video on splitting out Adaptive Processing Servers and did so in the test environment:
Its good and informative. However you will need to go into Monitoring and change the Server rules, the video does not really tell you that.
All the servers I created are at Failed(circle with white stripe) or have a red box. You need to go into Monitoring to change the rules.
To compound the issue I don’t have Adobe Flash, you need this for monitoring to work, in the test environment and need to get infrastructure team to do this.
I found out that we had a Support Account, “Nice”.
The Support People took the logs, created in the Logging Folder, and discovered one of the Services on the Adaptive Processing Server (APS) was causing the issue.
The APS Platform Search Service (PSS) has a locked table on the Database contained within the Central Management Server (CMS).
If I went into BI\BO Launch pad and tried to search for something I would get this message:
“An error occurred while searching: Index is not ready. Please try again later (PSS 01004). (PSS 01004)”
Nobody uses the Search function or if they did they did not report an issue.
Once the clone is created, Central Management Console (CMC) go to Applications, right click on Platform Search Application, select properties, click on ‘Stop Indexing’.
At the time of writing this I am hoping it resolves the issue.
Locked Table
Support are going to send me details on how to remove a lock on a table within the CMS database.
(BOB member since 2006-03-30)
(BOB member since 2002-08-15)
Well, I followed a very straight forward video on splitting out Adaptive Processing Servers and did so in the test environment:
