Having looked at the list archives, I see this issue has come up a couple of times. Reason I’m writing is I’ve seen the Weight parameter not being respected, often choosing the node I’d least want it to use when the cluster is almost entirely idle. I’ve seen one of the questions about this quoted the information about the Topology plugin and interactions:
NOTE:Slurm first identifies the network switches which provide the best fit for pending jobs and then selectes the nodes with the lowest "weight" within those switches. If optimizing resource selection by node weight is more important than optimizing network topology then do NOT use the topology/tree plugin. Here’s my question: how does the topology plugin determine which switch to use on an idle cluster? It seems to favor the exact nodes I don’t want it to favor, presently. Is there some way to at least make it choose other switches first or to order things in such a way that this is likely to happen? Is there any chance that it’s using alphabetical order or something? I’m running SLURM 15.08 currently, but planning to move to 16.05 (or perhaps 17.02) shortly. Thanks. -- ____ || \\UTGERS, |---------------------------*O*--------------------------- ||_// the State | Ryan Novosielski - [email protected] || \\ University | Sr. Technologist - 973/972.0922 (2x0922) ~*~ RBHS Campus || \\ of NJ | Office of Advanced Research Computing - MSB C630, Newark `'
