Remove redundant persist_rem field from job record
Fix logic for buffer state related to failure handling
Move pre_run logic into a separate thread for improved system performance
Disable changes to job burst buffer field
burst_buffer contents are validated at job submit time and
  data is possibly being staged at later times. It can not
   be changed.
Refactor bb_p_job_cancel() logic
Fix some job staging problems
Eliminate logging of unchanged pool information
Document that job arrays don't support burst buffers
Refactor job buffer allocation to eliminate race condition with
  stage out. Needs more work after prolog_running functionality
  is in place (merged from v14.11).
Improve cold-start state clean up

since you can share nodes.  Other TRES (as of writing this) aren't shared.

now nodes are TRES.

under TRES.

This can (hopefully) be used as an example when adding new TRES in the
future.

code.

Clean-up of of job stage-out and teardown logic
Improve restart state clean-up

in the database.

Fix some job state logic
Add logic to preserve persistent buffer account, QOS and partition for limit enforcement

and you are looking for idle nodes to make the partition the node you
run all your tests on will not be in the partition you create because the
node you ran on before the partition was created wasn't idle when you
tested it.

Fix slurm.spec file for recent layouts change
Refactor bb_pack_bufs() and bb_alloc_job_rec() functions

Add partition information to buffer record for limit enforcement

Add account and QOS to burst buffer data structures for TRES limit work
sview: Correct size units
sview: Fix free of uninitialized iterator logic
burst_buffer.conf: Fix parsing of DefaultPool parameter

Add support for access and type fields in interactive buffer creation

it most likely had bad limits set making it so we wouldn't be able to
run a job.

It turns out we need to send the update every time since there isn't a
super quick way to tell the association didn't change.