SchedulerParameters now accepts keyword of "max_rpc_count" in
  addition to the documented "max_rpc_cnt"

sched/backfill: If max_rpc_cnt is configured and the backlog of RPCs has
    not cleared after yielding locks, then continue to sleep.

requesting any specific association.

calls.

snowflake10 after snowflake8

Disable srun --bcast option test if SlurmdUser != root as the
  operation/test  will not work otherwise.

This would have helped to identify the performance issue associated
with bug 2174

sacctmgr output used a suffix of "PU" for per user, but only
  accepted input with a suffix of "PerUser". The output all
  used a suffix of "PU". The code has been modified to accept
  both sufficies (e.g. MaxJobsPerUser=# or MaxJobsPU=#).

bug 2147

BurstBuffer/cray: Fix job record purging if cancelled from pending state.
The problem can occur when the a burst buffer record was created for
the job in the plugin data structure, but no burst buffers were actually
allocated for it.
bug 2165

BurstBuffer/cray: Enable clearing of burst buffer string on completed job
    as a means of recovering from a failure mode.
Format is "scontrol update jobid=### burstbuffer=".
partial resolution of bug 2165

This is a backup in the event that slurmstepd fails in such a
  fashion that nothing gets logged. Added to help diagnose
  bug 2028

bug 2028

BurstBuffer/cray: Add logic to terminate dw_wlm_cli child processes at
    shutdown.
bug 2166

Previous logic required the buffer name to work
bug 2167

Distinguish between job and step IDs in sbcast (or srun --bcast)
  logs.

bug introduced in commit 06b54823

Added srun option of --bcast to move executable file to compute nodes

burst_buffer/cray: Support file staging when job lacks job-specific buffer
    (i.e. only persistent burst buffers).
bug 2113

No change to document contents, just reorded the fields documented.