Document that persistent burst buffers can not be created or destroyed using
    the salloc or srun --bb options.
bug 2404

Missed in b5bba34c

This patch builds upon commit 0c27392b,
adding calls to the new function so the slurmctld daemon will log
reasons for jobs failing to start.
bug 1828

Add new RPC (REQUEST_EVENT_LOG) so that slurmd and slurmstepd can log events
    through the slurmctld daemon.
bug 1828

on batch script completes.

code change.

The problem only exists on a subset of KNL models, those with 68 cores.
bug 2941

And make it a pointer on the job_record struct.

Lost after rebase.

- Only ask if cluster is being assigned to a different federation.
- Show that cluster is already part of federation and don't add cluster to
list of updates.
- Don't send updates if nothing to change (after finding that clusters are
already part of federations).

New jobs will get job IDs within the new range. Even though the
pre-existing large jobid-jobs will have federated jobids, they will be
treated as local jobs.

Fix by allocating slurm_msg_ts and freeing instead of using local
variables.

Display local cluster first and don't show persistconn.
$ scontrol show fed
Federation: f1
Self:       fed1:127.0.0.1:31002 ID:1 FedState:ACTIVE Weight:0
Sibling:    fed2:127.0.0.1:32002 ID:2 FedState:ACTIVE Weight:0 PersistConn:Connected
Sibling:    fed3:127.0.0.1:33002 ID:3 FedState:ACTIVE Weight:0 PersistConn:Connected

Appends federation related columns to default format options.