IPC Instability

[Former Member]

Hello,

We have recently upgraded from EBP 3.0 to SRM 4.0 (SRM Server 5.0). New to us is the IPC, of which we are using version 4.0, support pack 6. Periodically (approximately every 7-8 business days) the IPC stops working. This causes numerous short dumps in our system entitled UNCAUGHT_EXCEPTION. We are running on an IBM P570 server using an AIX level 5.3 operating system. We have adjusted AIX Java memory parameters, recycled the IPC services for said parameters to take effect, however we are still experiencing the sporadic IPC drop and subsequent dumps. If anyone has any insight on how to correct this problem and make the IPC more stable, I would greatly appreciate it.

Thank you kindly,

Kristy Barnes

[Former Member · Accepted solution]

Hi Kristy,

what do you mean by "stops working"? Does your IPC hang? Does it restart (e.g. because it runs out of memory)? Does it recover on its own or do you need to restart it?

Typically the IPC log files can show what's wrong at the time when it has problems.

There can be numerous reasons for your problems (hardware not sized properly and running IPC with too high load, sudden load increases because of scheduled batch jobs, restarting your SRM system, ...).

We've also had trouble with the AIX Java VM's garbage collection, which does things differently. You might want to take a look at OSS note 716927, which is not really IPC, but you can set up your Java VM quite similarly for IPC than for the SAP J2EE.

If you are experiencing serious issues, it's best to open an OSS message.

Best regards,

Karlheinz

[Former Member · Accepted solution]

Hi Karlheinz,

Thank you so much for your reply. I'll try to explain our situation a little bit better. We have the IPC installed on the Unix server that is hosting our SRM Server 5.0 Web Application Server (ABAP). After approximately 9 business days' worth of activity, the IPC just stops working, echoing an "Interpreter Failed" Error message to the end user. It won't properly function anymore, and the quickest resolution is to stop the IPC services and subsequently restart them again. Once complete, the users no longer encounter the error I mentioned above.

Thank you for mentioning the OSS note. We have looked at that one per guidance by SAP, and recently applied the initial Java heap size and maximum Java heap size memory parameters stated in the note. So far, the IPC has been fine, but it's probably too early to tell. We currently have a customer message opened with SAP and have established a lot of dialog back and forth regarding the matter. What I'm trying to find now is the -Xss argument for Java. I do not see this in any of the IPC files I've checked so far. I feel that if I allocate a bit more memory here we may be closer to a solution. The reason I state this is because in those log files you mentioned, I see repeatedly the following error message:

Fatal: Exception : JVMXE018: Insufficient Java stack size

java.lang.StackOverflowError: JVMXE018: Insufficient Java stack size

Again, thank you so much for taking the time to look at my message and respond. Your input is greatly appreciated!:)

Kind regards,

Kristy Barnes

[Former Member · Accepted solution]

Kristy,

Did you get satisfaction on this error ?

Fatal: Exception : JVMXE018: Insufficient Java stack size

java.lang.StackOverflowError: JVMXE018: Insufficient Java stack size

We are receiving this from a UserExit when we access a different type of Java servlet - Taxware. It is happening on large amounts of transactions.

Do you know where your error message JVMXE018 is coming from ? We are not sure if it's the communication channel or the Backend Taxware servlet.

[Former Member]

Hi Kristy Barnes,

We are also getting exactly same error as you faced before with your IPC. Do you remember the solution for your issue. If yes, please share the resolution. I have opened a message to SAP 2months back, but there is no proper answer till now.

Please note that we are facing this occasionally but not every 9hrs as it happened in your case. It is like once in a month or so.

Our environment is same as yours.

Thanks,

Giri