Aggregate job status polls into a single backend command #443
Description
It would be great if we could poll multiple pending tests using a single backend command such as follows:
sacct -j job1,job2,job3
squeue -j job1,job2,job3
This optimization will significantly drop the SLURM database hit rates and we could even allow us to have smaller polling intervals, because we could poll multiple job statuses at the same time.
However, the challenging part for this is designing correctly this concept, because it crosses the internal Job API. The framework sees jobs as individual entities and does not know how the submission or polling happens. It can collect job IDs or jobs, but it cannot issue an aggregate sacct or squeue command, because this is job-scheduler specific. So a new entity must lie in-between to abstract this aggregate operation.