Jira running out of memory when syncronizing user directory

Andrew Roesener

I am synchronizing several user directories in my Jira application as our company has multiple domains. I am running into an issue with one of them when it syncs it causes a Java error after eating up all available memory. The job has succeeded before and the filters only pull a handful of users and groups so I am not sure why the synchronization eats up so many resources.

Job:

com.atlassian.jira.crowd.embedded.JiraDirectoryPollerManager.10100

Error:

OperationFailedException: java.util.concurrent.ExecutionException: java.lang.OutOfMemoryError: GC overhead limit exceeded ExecutionException: java.lang.OutOfMemoryError: GC overhead limit exceeded OutOfMemoryError: GC overhead limit exceeded

Earl McCutcheon

Hi Andrew,

Garbage Collection (GC) memory errors are a delicate balancing act of application usage to resource allocation, and there is no easy answer for this error unfortunatly, as covered in the KB article covering GC errors "OutOfMemoryError GC Overhead Limit Exceeded" : 

more memory does not always equate to a better experience

But In your particular use case the system does seem to be getting overloaded during the sync event itself from your description of the events timing, and this being the case would benefit from a Bump in avaliable memory on your heap space, especially if this occurs regularly during sync events and not just a one off event.

Moving forward this is going to require a lot of performance monitoring to fine tune the settings to align with the usage levels and I would recomend looking at enabling Garbage Collection logging as covered in "Using Garbage Collection Logs to Analyze JIRA Application Performance" with a heavy focus on the Tuning guide pointed out at step 2 of the analysis section "Garbage Collection (GC) Tuning Guide"  for the principles to follow for setting the memory profile to handle the increased events leading you to OOME.

Regards,
Earl