killed even if the node where srun executes goes down (unless the
    --no-kill option is used, previous termination logic would fail if 
    srun was not responding).

    specific subsystems. This is just the infrastructure, flag is unused
    right now (work on tomorrow).

    (cons_res_preempt.patch from Chris Holmes, HP).

    the TaskPluginParam configuration parameter rather than slurmd's
    SLURM_ENFORCED_CPU_BIND environment variable.

  (also added output of SacctDefaultCommand from "scontrol show config")

  and Moe Jette. There are still a couple of end-cases that need to be fixed. We also
  need to push the core identification out to the slurmd for task binding. sched/gang
  needs revision to support this too.

    a job's completion before allocating already released resources to pending
    jobs. This can be used to reduce fragmentation of resources.

  documentation, read config, show config, and run PrologSlurmctld when a job
  is allocated resources (startup synchronization still needed).

    https://computing.llnl.gov/linux/slurm/power_save.html or 
    "man slurm.conf" (SuspendProgram and related parameters) for more 
    information.
This is the final installment of the work: update some documenation, 
    increase default ResumeRate, and reduce frequency of retrying 
    batch launch state check.

    "man slurm.conf" (SuspendProgram and related parameters) for more 
    information.
NOTE: the step create logic needs to be modify to return EAGAIN or the 
    like until the node's state NOT_RESPONDING flag gets cleared.

    longer needed.

See NEWS for identification of all changes

    ping the node immediately to clear the NOT_RESPONDING flag.

    expression rather than one line per node. Frequency of log messages is 
    dependent upon SlurmctldDebug value from 300 seconds at SlurmctldDebug<=3
    to 1 second at SlurmctldDebug>=5.

    the node to be returned to use when it starts responding rather than 
    remaining in a non-usable state.