Uploaded image for project: 'Confluence Data Center'
  1. Confluence Data Center
  2. CONFSERVER-3430

Confluence (or possibly Tomcat) seems to have a memory leak, possibly in relation to hanging on to session

    • Icon: Bug Bug
    • Resolution: Timed out
    • Icon: Medium Medium
    • None
    • 1.4.1
    • None
    • SunOS, jdk1.4.2, Tomcat 5.0.28

      One of our system admins pointed out to me yesterday that our instance of Tomcat (mainly running Confluence) was using 45% of the CPU, and ~500Meg of memory.(see screen shots) I think Tomcat had been running almost exactly a week without restart. We have roughly four frequent users of this system and half a dozen casual users at present. Notice too from the screen shots the 241 sessions the Confluence context was hanging on to. This should have been about 4 sessions with our standard timeout settings.

      Once I used the Tomcat manager to stop the Confluence context everything dropped to normal with CPU usage down to ~0.05% and sessions dissappearing from the Tomcat manager.

      I've noticed too that doing a reload of the Confluence context in Tomcat sometimes causes an 'out of memory' exception meaning Tomcat is likely not letting go of everything it should.

      The unusually high number of sessions noted in the Tomcat manager screenshot seems to at least be one starting point.

      Any advice on why CPU and memory usage is so high? Are there any known memory leaks?

      It is possible these are Tomcat issues and not Confluence issues. Are 'high usage' sites (such as Atlassian) finding you have to stop / start the context regularly?

        1. tomcatManagerScr.jpg
          tomcatManagerScr.jpg
          51 kB
        2. topScrShot.jpg
          topScrShot.jpg
          57 kB

              Unassigned Unassigned
              bpatters7 Brendan Patterson
              Affected customers:
              0 This affects my team
              Watchers:
              3 Start watching this issue

                Created:
                Updated:
                Resolved: