Change node_req field in struct job_resources from 8 to 32 bits so we can
run more than 256 jobs per node.

correctly as of IBM driver V1R1M1 efix 008.

Conflicts:
	src/common/read_config.c

configured correctly.

switch/nrt: un/pack job request for PMD use

switch/nrt: pass launch_state info to PMD

Specifically if memory allocations are not tracked, this code could
prevent jobs from running due to lack of memory.

Move code to handle multiple clusters.  Previous code could have issues
where job_db_inx space could overlap.

current or in the past.

from Mark Nelson

Using the syscalls directly rather than calling bin/(u)mount via
system() avoids a few fork + exec calls, and provides better error
handling if something goes wrong.

Users of this functionality are also updated to use slurm_strerror in
order to provide a more informative error message.

The mount and umount syscalls are Linux-specific, but so are cgroups
so no portability is lost.

Using the syscalls directly rather than calling bin/(u)mount via
system() avoids a few fork + exec calls, and provides better error
handling if something goes wrong.

Users of this functionality are also updated to use slurm_strerror in
order to provide a more informative error message.

The mount and umount syscalls are Linux-specific, but so are cgroups
so no portability is lost.

Using the syscalls directly rather than calling bin/(u)mount via
system() avoids a few fork + exec calls, and provides better error
handling if something goes wrong.

Users of this functionality are also updated to use slurm_strerror in
order to provide a more informative error message.

The mount and umount syscalls are Linux-specific, but so are cgroups
so no portability is lost.

    the current plugin has been loaded when using runjob_mux_refresh_config

DBD rpc version to SLURM protocol version.

could be called from a dbd pack/unpack as well as from regular rpc's.
This attempts to fix the issue buy passing a variable to the function to
adjust from the dbd rpc version to the slurm rpc version.

These comments were orphaned with this commit:  874f797f
Move the start time calculation of pending jobs into a separate pthread on Nov 3, 2010.

These comments were orphaned with this commit:  874f797f
Move the start time calculation of pending jobs into a separate pthread on Nov 3, 2010.

Code improvements for recording GRES use:

The xmalloc() function initialized everything to zero. There is no need for
a call to memset() afterwards.
The xfree() function tests for a NULl arguments and does nothing in that case.
There is no need for an "if" test before calling xfree().
There is an xrealloc() function available that eliminates the need for saving
and restoring values.
I made some formatting changes to comply with the Linux kernel coding standard,
mostly related to lines over 80 characters long.
JobID and some other fields should be formatted using "%u" rather than "%d"
to avoid negative job ID's being logged.
I added tests of buffer sizes in a couple of places to avoid possibly going
off the end of a buffer.
Code for the function gres_get_value_by_type() was missing from the patches.

Gres accounting modifications

Fix memory leak, avoid xstrdup().
Bump the rpc_version number in for the slurmdbd.
Break up some long lines.

gres accounting: save/restore job gres requested/allocated info, get job counts properly

gres accounting: updates based upon email from Steve Trofinoff, 31 July`

gres accounting: remove job gres_res and just replace gres

already open File * to write logs to.  This is helpful when something
sets up logging outside of SLURM but wants SLURM debug written to the same
place as the rest of there logs.

This particularly effects --multi-slurmd mode and the Port option