Java Engine is slowing down very frequently.. How to stop this.!!

[Former Member]

Hi Guys,

Java engine is becoming very slow sometimes in our production server and it is leading to connectionl lost between ABAP and Java engines. Once we restart the Java engine, the system will run normally and messages will process smoothly.

We couldn't able to find the reasons why it is happening. It is becoming a burden on us as it will happen sometimes during night times and weekends.

- Many times it is happening when the Java garbage collector runs for long duration.

- We are getting alerts only when the entries in SMQ2 is increasing and reaches to some point ( we set to 3000 entries) due to java down.

Is there any reasons why it is happening?

Is there any CCMS alert configuration we can do, so that we can atleast alerts when the java engine is slow?

Please suggest what might be the reason.

Thank you

Deepthi.

[agasthuri_doss]

Hi,

What is the system memory, Is it > 3 GB per Java Node.

Cheers

August

[Former Member]

Hi ,

>> please provide your OS/DB/JDK/SP stack version

OS : AIX

DB: Oracle

SP: 15

JDK.Version : 1.3

mak.real : 645_val_rel

>> We have the same issue sometime back, after additional java node implementation issue is resolved.

Is Java nodes are different to server nodes? We already have 3 servernodes running.

We done few changes as per SAP recommendation.

1. As most of our scenarios are based on File and SOAP adapters, We increased the "Maximum Number of Threads" to 10 instead of default 5 for these adapters.

SOAP_http://sap.com/xi/XI/SystemSend

File_http://sap.com/xi/XI/SystemSend

2. SXMB_ADM > Integration Engine Configuration > Specific Configuration > SA_COMM - CHECK_FOR_MAX_SYNC_CALLS = 200

3. we set the below parameter in all the sender communication channels. (Note: 801926)

Advcance mode-> Parameter name: "clusterSyncMode" Parameter values: "lock"

On an average around 65,000 messages (both outbound and Inbound) will be processing every day. And we kept 14 days retention period for archiving and deletion of the messages.

When this issue happened, we check the thread dumps and found that out of 500 application

threads, there is 0 free thread.

On Server0, 208 application threads are blocked by thread:

"XI File2XI[CC_File_Sender_OrdersFile/BS1/]_63734"

On Server1, all running application threads are used by "File2XI".

On Server2, I see that there are only 90 application threads.

>> Is it > 3 GB per Java Node.

15GB alloted for total system and 2GB alloted for each server node.

Any suggestions please ??

Thanks

Deepthi

[Former Member]

Did you follow note 723909 to configure JVM?

Besides, importantly

> JDK.Version : 1.3

Is this version what you get from java -version? If so, this is not the recommended JVM for SAP Java application from performance perspective, you should go with IBM J9 JVM.

note 1024700 - Installation/Upgrade information for NW04/NW 7.0 on AIX

note 1008311 - Recommended Settings for NW 7.0 >= SR2 for the AIX JVM (J9)

Note 716927 - Overview of AIX JVM for NetWeaver 2004 and 7.0 (2004s)

Regards,

[Former Member]

Hi Deepthi,

Generally Java stack performance degrades based on multiple reasons like high volume of messages at that instant, improper java programs leading to memory holes/loops, JDK version associated with Operating system etc. But the common and simple solution for Java performance degradation is, increase the Java nodes in ur system to balance the load on java stack.

We have the same issue sometime back, after additional java node implementation issue is resolved.

Following notes will be helpful

SAP NOTE 1016241 - J2EE Engine hangs because of unsufficient system threads

SAP note: 716604 and upgrade JDK.

SAP note: 723909 -> set correct VM parameters

SAP note: 1259465: Memory analyser.

Thanks,

RJ

[Former Member]

Hi my friend

This is a fuzzy topic, please provide your OS/DB/JDK/SP stack version info as detailed as possible for troubleshooting. Also, did it suddenly happen recently? Or happened after you applied or changed any configuration? Please do recall carefully all possible changes.

Regards,

Effan