force job to start like qrum/brun/... with other schedulers
dynamic node information like RANK or CODE or DATE and TIME set up in LSF
license information is very different than LSF, particularly per node and per cluster values
see who is doing queries like LSF querylog. But can limit number of queries
set limit on number of cores (and/or jobs) in general per user, not just for job arrays
custom account log and plots (via prologue and epilogue?)
explain terms for jobs,tasks,steps,reservation,active sibling job,federated job
explain use cases for reservation, step, task, job, ....
scrontab? is that new?
sview? reservations new
default is exclusive use of a node per job, which is horrible for single-core jobs. See consumable resources
would be nice if could easily just show single variable values from things like 'scontrol show config' like
>$ scontrol show config=SlurmctldHost,
> Slurmctld(primary) at mercury is UP
most schedulers create a file with hostnames in it for use as an MPI machine file or to create one. Do not see one for SLURM. Could use
sinfo --jobs=$SLURM_JOB_ID --format '%n' > machine_file