is still not running.
srun to wait for job initiation as needed with polling.

is still not running.

same processor count (as used for scheduling, configured or actual).
This will be required until the Elan interface is more sophisticated.

step_mgr was incrementing step_id twice for each job step started.
srun only notes that a job is complete when it makes the allocation,
otherwise it only notes that the job step has completed and the
allocation remains active.

step_mgr was incrementing step_id twice for each job step started.

Srun uses existing allocation if SLURM_JOBID environment variable
is set.

slurm_confirm_allocation.

available data).

(block or cyclic).

set, nodes will change state from DOWN to ALLOCATED or IDLE upon registration.

Add new srun options: --kill-off (no kill on node failure), --share, and
--time=minutes.