Hello, Please forgive me if this is something I could have gleaned from the documentation--I am new to SLURM administration. I work at a facility which has a large general-purpose compute cluster with ~20K cores that uses SLURM as the scheduler. My group has an allocation on this system, and we also have our own smaller cluster with about 700 cores that we control. My question is: is it possible to transparently link a SLURM job queue on our small cluster with our allocation on the large cluster? Ideally, when no open slots are available on the small cluster, we would like further jobs to go to our allocation on the large cluster, and then when both are full, load-balance the job queue between the two intelligently somehow. Also, we would want this to be a one-way relationship--we don't want any jobs submitted to the large cluster running on our small cluster.
We have a good relationship with the administrator of the large cluster, so I think he would be willing to do at least a little bit of work on his side to allow this, if it is possible at all. I don't even know the right terminology for what I am talking about, so any help is appreciated. Thanks in advance!
