Split a slurmctld's job record "shared" field into "share_res"
(share resource) and "whole_node" fields. Needed to better manage
allocation of whole nodes for core specialization without disabling
gang scheduling of such jobs.

continue the operation for other elements.

Signed-off-by: Danny Auble <da@schedmd.com>

to suspend/resume array elements.

This reverts commit 34fd501c.

connect in a loop instead of producing a fatal.

performance counters.

to get from the plugin available nodes to run step on.

This is needed for a Cray when running with Network performance counters.

slurmd.  This will make the node unresponsive if any communication error
happens.  When the slurmd comes in with the current newer version the
state is removed and all is well.  Previously if this state was around for
too long then the node would be marked down and kill running jobs.

This reverts commit 2fa28eb6.

Conflicts:

	NEWS

No change in text, just bad allignments

Conflicts:
	src/plugins/acct_gather_filesystem/lustre/acct_gather_filesystem_lustre.c
	src/srun/libsrun/opt.c

Add a configuration parameter and job_submit plugin that can
rate at which any single user can submit jobs to Slurm. Note that
a job array is counted as a single job.
bug 421

data correctly accumulating differences between sampling intervals.
Fix the data structure mismatch between acct_gather_filesystem_lustre.c
and slurm_jobacct_gather.h which caused the hdf5 plugin to log incorrect
data.

This augments an earlier patch that picked up the protocol version
number, which lets us distinguish between 2.6.0 and 2.6.4 instead
of just being able to tell it is version 2.6.x.