Greg,There should be no lasting harm done by interrupting communication between the schedd and negotiator.
Condor groups similar jobs together to make matchmaking more efficient. When you say that the negotiator is "spinning" on a job, do you mean that it only considers that job and not other jobs in the queue? If the other jobs in the queue have the same AutoClusterId in their ClassAd, then the "spinning" that you describe is normal behavior.
Perhaps condor_q -analyze can help you figure out why the job doesn't match to a machine.
--Dan On 7/20/11 4:18 PM, Gregory Skelton wrote:
Hi, It seems that in the process of taking my condor pool down yesterday, one of my submit machine wasn't finished talking to the negotiator. Now my condor jobs are not being matched to any of the worker nodes even though yesterday those jobs were running just fine. The negotiator just keeps spinning on 1 job, even if I put that job on hold it spins on the next job. Has anyone seen this before? Many thanks, Greg