One of our testers created an illegal topology.conf file.

He has a config you probably wouldn't see in production, but can see in
testing when you are sometimes given a collection of miscellaneous
resources.

          |-- nodes
switch1 --|
          |-- switch2 -- nodes

He tried the topology.conf file below. Switch s1 is defined twice. Slurm
accepted this config, but wouldn't allocate nodes from both switches to
one job.

SwitchName=s1 Nodes=xna[14-26]
SwitchName=s2 Nodes=xna[41-43]
SwitchName=s1 Switches=s2

I believe slurm shouldn't allow the second definition of switch s1. The
attached patch checks for duplicate switch names.
Patch from Rod Schultz, Bull.

This reverts commit 350ef5dc.

next parallel step is ran on a sub block, SLURM won't over
subscribe cnodes.

BLUEGENE - Modify "scontrol show step" to show  I/O nodes (BGL and BGP) or
c-nodes (BGQ) allocated to each step. Change field name from "Nodes=" to
"BP_List=".

cannot fit into the available shape.

Previous code would fail when trying to launch more than 4096 tasks,
which is a problem on BGQ systems where SLURM actually launches job
steps.

or not.

This change applies only to Cray systems and only when the srun
wrapper for aprun. Map --exclusive to -F exclusive and --share to
-F share. Note this does not consider the partition's Shared
configuration, so it is an imperfect mapping of options.

A node DOWN to ALPS will be marked DOWN to SLURM only after reaching
SlurmdTimeout. In the interim, the node state will be NO_RESPOND. This
change makes behavior makes SLURM handling of the node DOWN state more
consistent with ALPS. This change effects only Cray systems.

Fix the node state accounting to be consistent with the node state
set by ALPS.

be the same.

previously marked down by alps.

Require SchedulerTimeSlice configuration parameter to be at least 5 seconds
to avoid thrashing slurmd daemon.
Addresses Cray bug 774692

Change in GRES behavior for job steps: A job step's default generic
    resource allocation will be set to that of the job. If a job step's --gres
    value is set to "none" then none of the generic resources which have been
    allocated to the job will be allocated to the job step.
Add srun environment value of SLURM_STEP_GRES to set default --gres value
    for a job step.

On Bluegene systems, smap's command-line mode would generate an invalid
memory reference due to an uninitialized variable.

a POLLERR the dbd_fail callback is called.

the DBD where both remained up but were disconnected the slurmctld would
get registered again with the DBD.

With sched/wiki or sched/wiki2 (Maui or Moab scheduler), insure that a
requeued job's priority is reset to zero.

Add the ability for a user to limit the number of leaf switches in a job's
allocation using the --switch option of salloc, sbatch and srun. There is
also a new SchedulerParameters value of max_switch_wait, which a SLURM
administrator can used to set a maximum job delay and prevent a user job
from blocking lower priority jobs for too long. Based on work by Rod
Schultz, Bull.

BlueGene: Permit users to specify a separate connection type for each
dimension (e.g. "--conn-type=torus,mesh,torus").

On Cray systems with the srun2aprun wrapper, build an srun man page
that describes which options are available with the wrapper.

Restore node configuration information (CPUs, memory, etc.) for powered
down when slurmctld daemon restarts rather than waiting for the node to be
restored to service and getting the information from the node (NOTE: Only
relevent if FastSchedule=0).

Fix bug in select/cons_res task distribution logic when tasks-per-node=0.
Eliminates misleading slurmctld message
"error:  cons_res: _compute_c_b_task_dist oversubscribe."
This problem was introduced in SLURM version 2.2.5 in order to fix
a task distribution problem when cpus_per_task=0. Patch from Rod Schultz, Bull.

Set SLURM_MEM_PER_CPU or SLURM_MEM_PER_NODE environment variables for both
interactive (salloc) and batch jobs if the job has a memory limit. For Cray
systems also set CRAY_AUTO_APRUN_OPTIONS environment variable with the
memory limit.

For front-end configurations (Cray and IBM BlueGene), bind each batch job to
a unique CPU to limit the damage which a single job can cause. Previously any
single job could use all CPUs causing problems for other jobs or system
daemons. This addresses a problem reported by Steve Trofinoff, CSCS.

man pages. Patch by Nancy Kritkausky, Bull.

Bill Brophy, Bull.

Note the job and partition state file formats have changed and RPCs
with information for jobs and partitions have changed.

Fix bug in generic resource tracking of gres associated with specific CPUs.
Resources were being over-allocated.

Fix memory buffering bug if a AllowGroups parameter of a partition has 100
or more users. Patch by Andriy Grytsenko (Massive Solutions Limited).

Add cgroup support for device files in both the task/cgroup plugin and generic
resource (GRES) logic. Based upon patch Yiannis Georgiou.

When suspending a job, wait 2 seconds instead of 1 second between sending
SIGTSTP and SIGSTOP. Some MPI implementation were not stopping within the
1 second delay.

Add contribs/arrayrun tool providing support for job arrays. Contributed by
Bjørn-Helge Mevik, University of Oslo. NOTE: Not currently packaged as RPM
and manual file editing is required.

If a job needed to preempt other jobs to start and those jobs were
not completed by the time of the next scheduling cycle, other jobs
might be selected for preemption in that next cycle resulting in
more jobs being preempted than necessary.