state be NODE_FAIL rather than CANCELLED.

Based on 03_scontrol_clusters_argu~1.diff from Gerrit.

02__hostlist_width_equiv.diff from Gerrit.

01_hostlist_system_dimens~1.diff from Gerrit

    tasks.

    a job and let the job's owner release it. The scontrol command of
    "hold <job_id>" when executed by a SLURM administrator can only be released
    by a SLURM administrator and not the job owner.

Apply rlimits right before execing the users task so to lower the risk of the task exiting because the slurmstepd ran over a limit (log file size, etc...)

    is exceeded.

    slurmctld is reconfigured while a job is in completing state.

  The call will be issued from read_slurm_conf().

resources on the node, job suspend has no effect on allocated gres (previously
it did).

Fix for possible deadlock in the slurmstepd when cancelling a job that is also writing a large amount of data to stderr.

    CEA.

    queue of jobs rather than restarting at the top if there are no changes in
    job, node, or partition state between runs. Patch from Hongjia Cao, NUDT.

    (5 minutes) instead of one minute.

    terminated on. Patch from Hongjia Cao, NUDT.

Fix for problems when adding a user for the first time to a new cluster with a 2.1 sacctmgr without specifying a default account.