after slurm executes scontrol reconfigure, it tries to start (review states) all its nodes, but then we have errors in slurm's log like slurm can not access the nodes. How can I set slurm that he would put all its nodes to its an internal tables which he uses to power_save resources?
2015-01-22 22:09 GMT+02:00 Anatoliy Kovalenko <[email protected]>: > One more question: does slurm have a command which helps to remove a > node with status IDLE#+CLOUD or DOWN#+CLOUD from node list? The problem is > that slurm does not execute ResumeProgram when it sends task, > > 2015-01-22 17:23 GMT+02:00 Anatoliy Kovalenko <[email protected]>: > >> I setup a power manager (elastic nodes) for slurm >> http://schedmd.com/slurmdocs/elastic_computing.html but when I restart a >> slurmctld, it executes ResumeProgram immediately for all its nodes, which >> has CLOUD state in conf. Why this behavior happens? If I make change to >> slurm in nodename and make again restart slurmctld - everything is ok. But >> after a while this issue is come back. >> > >
