No more memory on JIRA

[Emmanuel Rouillard]

Our jira server (4.4.5) is for about 1000 users a month, 100 daily with 154629 issues but crashes sometimes because of not enough memory.
The memory allocated to JVM is more than 5Go : (VM arguments are -Xms5124m -Xmx5124m).
But sometimes the server is not responding because of the garbage collector that is conumming all the cpu time to clean the memory. Few seconds later, it runs again and agin. No user can access to JIRA during that time and we have to restart jira.
I think that memory is enough for our type of server, don't you think ?
How to know if it is a problem of jira or a problem of a third party plugin ?

Could you help me to find the reason why?

[Emmanuel Rouillard · Accepted solution]

Thanks for your replies. I increase the Heap Memory of JIRA. Now we have 7 Go of total Memory. It seems to correct the problem. According to Atlassian, so much memory is not required....

We have about 1500 users a month and 150 users a day that connect to jira at least once.

But we have many contexts and projects (300+)...

What do you think ?

What is your experience in this case ?

[Timothy]

Do a thread dump when it runs out of memory and you should be able to find out where it fails. The best way to improve performance IMHO? Upgrade your JIRA version. Version 4 is no longer supported too.

[Emmanuel Rouillard]

I can not make a thread dump because when GC is running, Jira is not responding. Impossible to get the monitor page from Melodymonitor, impossible de generate a thread dump. Have you got any solution ?

[Andris Grinbergs]

If you didn't tried it already - try setting PermSize=512m. Also you should check if you are running in 64 bit OS.

[Zul NS [Atlassian]]

154629 issues with -Xms5124m -Xmx5124m is too much. Could be the reason why GC running longer than it should. How long did the Full GC took in the GC log?

Recommend to lower the initial and max memory allocation to 2GB, should be sufficient for more than 200,000 issues. Also to look into the JIRA Sizing Guide, there is a Hardward Recommendation of Sizing Guide that is really helpful https://confluence.atlassian.com/display/ATLAS/JIRA+Sizing+Guide

[Emmanuel Rouillard]

When jira is not responding, I can see that the Full GC runs every line in the GC logs. I tooks sometimes 10 seconds but it did not recover very much memory. So it must run again few seconds later. I hardly believe that if I decrease memory the pb will disapear.... On the contrary, I think that pb may occur more often. It may come from a plugin but which one ???

Anyway thank you for your reply and to try to help me to stop the mess with Jira.

[Zul NS [Atlassian]]

If full GC running every lines of 10 seconds each, then this could be caused by other things. I don't know of your exact configurations at the moment, hence would not be as helpful. But I'd like to just confirms first if it is indeed comes from a plugin, can we enable safe mode?