External event handler services randomly stop for no apparent reason

[Former Member]

On Netweaver IdM 7.1, I set up a couple of simple event agents to detect file changes in order to execute an IdM job.  These work perfectly, just as expected, except that the services randomly stop on the server on their own.  I've had our Basis people investigate and they cannot determine a root cause. 

We have tried to configure the services so that they would restart themselves automatically when they go down, but this has not been successful.  I've never had any issues with dispatcher services, which are set up in much the same way.  Has anyone tried to use these "file change" event handler services and had similar issues?

Thanks,

Keith

[former_member2987 · Accepted solution]

Hi Keith,

Not 100% sure but I do know there's been a reported issue about dispatchers not coming back on line automatically.

One thing you might want to do is start the event agent in TEST mode, this will open a Command Prompt window that shows everything the service is doing.

One place I would look about sudden shutdowns is your Java settings.  Is it possible that you are running out of memory.  Again, the TEST mode should give you a better idea.

Hope this helps!

Matt

[Former Member · Accepted solution]

UPDATE:

After running the "file watcher" event agent/service in TEST mode, I noticed that multiple instances of the process were running.  This event agent is supposed to continuously check for changes to a particular file and execute a job if the file has changed.  I had the job deleting the file as a final step, meaning that the file did not exist for the event agent to find unless a new one was created and placed in the target directory.  Apparently, when the event agent does not find the file at all, it results in an error condition that triggers execution of an additional instance of the JAVA process.  in a short time, there are dozens of instances running and the memory is exhausted. 

I changed the job to leave the file in the directory after the job completes, and the event agent no longer times out or runs out of memory.  The only down side to this is that when IdM is bounced or the event agents are restarted for any reason, the agent WILL process the file again as if it's new.  In our case that shouldn't matter, but to others it might. Just be warned if you use this type of event agent.  In my opinion, SAP should code the JAVA in a way that more elegantly handles the "file does not exist" scenario without invoking another JAVA instance each time. 

[former_member2987]

Great to hear, Keith!

Maybe when the job finishes you should just create the file as a 0 byte file, or just the headers to take care of that remaining issues.

Regards,

Matt

[Former Member]

Matt - I thought about that and may end up going that route. Just wanted to find a solution that also didn't kick of the job unnecessarily.  Even if I clear the file or create a blank file, the job will execute, and I may run the risk of interfering with new valid data being pushed.....but that is a valid option.

[former_member2987]

Keith,

One of the many reasons I don't like Event Agents... 

Matt