Hello again,

Here's what my system can do using the multi-cluster functionality:
- sbatch searches for the less loaded cluster (
*sbatch_set_first_avail_cluster* function in sbatch.c).
- Job is sent to the chosen cluster.

But, when the job is in the queue of a cluster, it cannot be moved to the
second cluster even if it is free.

Once the job id is set, it's unique within a cluster and cannot be
modified. I guess moving this job to another cluster could result in a
unicity problem if we keep the same job id.

So I tried to analyse *sbatch.c* in order to see how and when the job id is
assigned. Untill now I haven't figured out when this is done.


Can someone tell me when the job id is assigned and which function/program
is in charge of this ?

Reply via email to