On Wednesday, May 4, 2011 at 3:53 PM, Larry Owen wrote:
Thanks for the help earlier. New issue, or should I say Newb issue.
I have 2 nodes in my test cluster, one running all the services the other just master and startd.
condor_status will show both, however when I submit a jobs they never run on the worker. condor_status -available will list the node as available.
I looked at Negotiator.log, but can't find any clues. How do I go about finding the cause?
Check the job's requirements attribute:
condor_q -f "%s\n" requirements
Do they constrain you out of running on your other machine?
You could consider shutting off condor_startd on the good machine (condor_off -subsys STARTD) so that the job only has the one bad machine to run on. And once you have a job in the queue, idling, try running:
condor_q -better-analyze <cluster>.<proc>
Where <cluster> and <proc> are replaced by the cluster ID and process ID of your idle job. This command will give you an analysis of your job's requirements _expression_ and can make suggestions on what to change to get it to match machines.
If the output isn't clear, cut and paste it to this thread and one of us can give you additional guidance.