Running a service when no other jobs are available

• Question

• 

  

  0

  Sign in to vote

  I have 4 nodes in my cluster that I want to devote to distributed code builds when there is nothing else to run.

  The way I expected this to work was that I would start a job that has service tasks with the lowest priority, and they would just run on every available node. Unfortunately, I have to keep an eye on it because there are a few problems:

  •  When I start the job, only the nodes that are online at that time get jobs (nodes that come online later don't get a task). I have it specified by node group.
  •  When a node loses network connection after a certain time it seems to give up on it and not reconnect

  To me it looks like when I submit the job all the tasks are calculated once based on the state of the cluster, where I want it to continually add or refresh the tasks. Is there any way to do what I want?

  Friday, October 5, 2012 9:32 AM