timeout when talking to the Database is the same timeout so a race
condition could occur in the requesting client when receiving the response
if the database is unresponsive.

the entire cnode.

Fix SelectTypeParameters=CR_PACK_NODES for srun making both job and step
resource allocation. Previously a stand-long srun command would distribute
tasks evenly across the nodes.

Start a job in the highest priority partition possible, even if it requires
preempting other jobs and delaying initiation, rather than using a lower
priority partition. Previous logic would preempt lower priority jobs, but
then might start the job in a lower priority partition and not use the resources
released by the preempted jobs.
bug 1032

together.

line buffered.

Job array dependency based upon state is now dependent upon the state of
the array as a whole (e.g. afterok requires ALL tasks to complete
sucessfully, afternotok is true if ANY tasks does not complete successfully,
and after requires all tasks to at least be started).

Count is reset running array tasks on reconfig

sched/backfill - Set expected start time of job submitted to multiple
partitions to the earliest start time on any of the partitions.
Previous logic would set the time to that of the last partition
tested.

Previously job would only run in first listed partition.

Fix gang scheduling for jobs submitted to multiple partitions.
Previous logic assumed the job's "partition" field contained a
single partition name, that in which the job is running. That
was recently changed in order to support job's being requeued,
which we want to be runable in all of it's valid partitions.

Added squeue -P/--priority option that can be used to display pending jobs
in the same order as used by the Slurm scheduler even if jobs are submitted
to multiple partitions (job is reported once per usable partition).

instead of guessing off the exit_code.

signal 1.

Modify crypto/munge plugin to use socket and timeout specified in AuthInfo.

done for normal steps.

Modify AuthInfo configuration parameter to accept credential lifetime
and socket path options. Previously it accepted a socket path only.

of acting like it is a signal and exitcode.

signal only the steps and unless, in the case, of a batch job B is
specified in which case signal only the batch script.

previous it was always -2.

Add node state string suffix of "$" to identify nodes in maintenance
reservation or scheduled for reboot. This applies to scontrol, sinfo,
and sview commands.
Enable scontrol to clear a nodes's scheduled reboot by setting its state
to "RESUME".

Apply BatchStartTimeout configuration to task launch and avoid aborting
srun commands due to long running Prolog scripts.
bug 978

When nodes scheduled for reboot, set state to DOWN rather than FUTURE so
they are still visible to sinfo. State set to IDLE after reboot completes.
bug 1007

Srun executable names beginning with "." will be resolved based upon the
working directory and path on the compute node rather than the submit node.

Always call select_g_step_finish() when terminating a job step,
even if the job is also being terminated. This is needed for Cray
systems.
bug 1012

Fix race condition in CPU frequency set with job preemption.
When the preemptor job completed, it would notify the srun, which
would notify the slurmctld, which could resume a preempted job.
That preempted job could reset the CPU frequency before the
preemptor. This change has the slurmstepd resetting a job's
CPU frequency prior to notifying srun of completion, which
eliminates the race condition.
bug 1011