Just wait for the node to register with Slurm. If it doesn't register
within ResumeTimeout, the node will be marked down.

Continuation of d42889fb

Bug 6587

What was happening in a maintenance reservation was every 5 seconds
the slurmctld was sending a message to the slurmdbd telling it the
node's down/drain state was in a maintenance state as well.  We really
only needed to know about this once.

Bug 6487

(DRAIN+MAINT+IDLE) Before it would just print MAINT
Bug 6487

Bug 6487

Bug 4296

This will be used in future commits.

Bug 4296

GrpTRES behavior was changed with commits 7d69a43a and 158b88de so
that each job allocation is no longer counted separately towards the
GrpTRES limit. This is important for the GrpNodes.

Bug 5303.

to avoid having to send alias list for cloud nodes that exist in DNS.
This is especially important for large cloud environments (e.g.
thousands of nodes) as the alias list environment will be too large for
execve().

Bug 6589

Fix some logic committed in 6705b63c that caused test7.17 to fail

Bug 6587

This is useful in a cloud environment where the nodes come and go out of
DNS.

Bug 6592

bug 6445

Bug 6519.

Continuation of 324404de

Bug 6433

Continuation of c2cdde85

Bug 6433

If a GrpNodes limit is configurated in an association, partition QOS or
job QOS then favor use of nodes already allocated to that entity. This
will result in the configured node "Weight" being incremented by one for
nodes which are not prefered. Consider adjusting configured node "Weight"
values to achieve the desired node preferences.
bug 5303

Same as salloc/sbatch --gres option.

Bug 6582.

Bug 6387.

When memory.soft_limit_in_bytes is higher than memory.limit_in_bytes,
equal the soft to the hard, log a debug2 message and change
AllowRAMSpace documentation accordingly.

Bug 6071.

This reverts commit c79fd9ea.

Bad author.

When memory.soft_limit_in_bytes is higher than memory.limit_in_bytes,
equal the soft to the hard, log a debug2 message and change
AllowRAMSpace documentation accordingly.

Bug 6071.

allow_accounts, allow_alloc_nodes, allow_groups, alternate,
billing_weights_str, deny_accounts, deny_qos.

Bug 6595.

Handling now present in salloc and srun. This will be used for X11
forwarding, but can also work for other network traffic if desired.

Otherwise srun and other commands that load all select plugins
will fatal() with CR_ONE_TASK_PER_CORE set.

Disable a configuration of SelectTypeParameters=CR_ONE_TASK_PER_CORE with
SelectType=select/cons_tres. This will be addressed later.

salloc --no-shell jobs aren't interactive, and won't respond to pings
from the slurmctld, so they were previously getting killed after
InactiveLimit seconds, even when they had a running step. InactiveLimit
no longer applies for salloc --no-shell jobs.

Bug 6560.

Bug 5900.

Bug 5520

This value must be set to turn on detection of the library.
Without it the GPU GRES will default to previous Slurm behavior.

Bug 5520

Not much new code written, just moved around a bit.

bug 5520