How Do I Identify The Cause Of Garbage Collection Issues?

[Dalectric]

Hi

I've been battling this issue for some time now, but basically our Confluence instance will be stable and happy running for many months at a time with no changes to the JVM memory size and then suddenly I get a problem where the GC cannot recover enough memory. This makes Confluence inaccessible for all users as it doesn't respond.

From the GC logs I cannot see a trend in the memory usage that would warn me this is about to happen, it just suddenly stops being able to recover enough memory.

I have increased the maximum memory pool for the Confluence service so I'm now at 3840 MB, but as I increase this it obviously creates a bigger load on the server when it hits the limit again.

1. How can I identify what is causing the sudden increase in memory use?
2. How can I give myself more warning that I'm getting close to my memory limit?

Some example data below

 

[Dalectric]

I found the Java DK JVisualVM connected to our test server really helpful for identifying the cause of the problem. See below.

[Dalectric]

This has been identified to be caused by the File Conversion plugin, which has a memory leak and causes the available Java Heap space to fill up and then kills our server as it tries to recover it on GC.

We've tried up to 10GB Heap and this is still an issue so we've disabled the File Conversion plugin until this is fixed.

https://jira.atlassian.com/browse/CONFSERVER-46542

No idea when this will be fixed, but disabling the plugin reduces the functionality of Confluence for us, we were relying on this preview for collaborating on documents, or just easy preview of embedded documents.

https://confluence.atlassian.com/confkb/file-conversion-service-causes-out-of-memory-errors-in-confluence-5-7-or-above-703398523.html