JPPF Issue Tracker
Please log in to bookmark issues
CLOSED  Bug report JPPF-440  -  Nodethreads load-balancing algorithm not working as expected
Posted Feb 15, 2016 - updated Mar 21, 2019
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 5.1.x
Issue description
From this forum thread.

I've noticed during my tests that the nodethreads algorithm does not distribute the tasks according to the number of processing threads in the node. With the configuration discribed int he reproductionsteps, the expected tasks distribution is 12-6-3-3 but i get something very different: 10-6-4-4 in my latest test for instance. I added log traces and could observe that the algorithm reports a number of tasks to send that has nothing to do with the corresponding node's number of processing threads

Steps to reproduce this issue
Start 1 driver, 4 nodes with 4-2-1-1 processing threads respectively, then start an admin console, then submit a job with 24 tasks that just sleep for 10s each. You can observe in the admin console that the distribution of executed tasks to the nodes is wrong.

Comment posted by
Feb 15, 21:07
In fact this problem only happens if the number of processing thread in one or more of the nodes is changed after the node is connected with the server. The number of threads can be changed remotely either from the administration console or via a JMX API call. The cause of this is that the server does not update the number of procssing threads of its nodes, and the nodes do not notify the server when ever their number of threads is changed.
Comment posted by
Feb 17, 09:14
Fixed in: