JPPF Issue Tracker
Please log in to bookmark issues
CLOSED  Bug report JPPF-87  -  Memory leak in the driver: ServerJob instances fill the heap
Posted Oct 22, 2012 - updated Oct 23, 2012
icon_info.png This issue has been closed with status "Closed" and resolution "RESOLVED".
Issue details
  • Type of issue
    Bug report
  • Status
  • Assigned to
  • Progress
  • Type of bug
    Not triaged
  • Likelihood
    Not triaged
  • Effect
    Not triaged
  • Posted by
  • Owned by
    Not owned by anyone
  • Category
  • Resolution
  • Priority
  • Reproducability
  • Severity
  • Targetted for
    icon_milestones.png JPPF 3.2
Issue description
When performing some load testing, I got an OutOfMemoryError in the driver. I started to analyze the resulting heap dump and found that 648 instances of ServerJob were retaining 471 MB of heap (out of 512 MB max) I am attaching a leak suspects report generateed from Eclipse MAT.
Steps to reproduce this issue
  • use 1 driver: -Xmx=512m, load-balancing = proportional
  • use 1 node (non local, but I presume reproduction would occurs with a local node too)
  • run the matrix multiplication (300x300) sample for 1000 iteration, client load-balancing = default, 1 driver connection
  • ==> you get the OOME after about 650 iterations

Comment posted by
Oct 22, 07:34
A file was uploaded. leak suspects reporticon_open_new.png This comment was attached:

To view the report, unzip the file, and browse from the "index.html" page. If you click on the "details" link, the "accumulated objects per class" section clearly shows that there are 648 instances of "ServerJob" in the JPPFPriorityQueue, which retain around 470 MB of heap (I configured with -Xmx512m).
Comment posted by
Oct 23, 13:08
Fixed in trunk revision 2500

The issue was updated with the following change(s):
  • This issue has been closed
  • The status has been updated, from New to Closed.
  • This issue's progression has been updated to 100 percent completed.
  • The resolution has been updated, from Not determined to RESOLVED.
  • Information about the user working on this issue has been changed, from jandam to Not being worked on.