responding in a timely fashion) then send a registration RPC
to slurmctld.

are allocated. Replace polling to test for allocation with processing of
allocation response RPC from slurmctld at allocation time.

   to specified user.
 o New srun option --no-shell
 o Changed uid/gid_from_name () to uid/gid_from_string () and allow
   either user/group name or numerical user/group ID.
 o updated srun(1) documentation with new --no-shell option and
   changed semantics for --uid.

registration (in particular after changing state from UNKNOWN to DRAINING,
an inactive node was not registering to reset its state to DRAINED)

   of slurmd process.

    calls, improves performance and eliminates realloc failure on
    RH EL3 under extremely heavy workload apparently due to memory
    fragmentation

Requires setting counts in updated batch job launch request message.

    corresponding file descriptor for readability. Otherwise, srun
    and slurmd will spin when stdin file descriptors are full because
    pending data will always go unread. (gnats:434)

specifying an infinite time limit.

are specific to a particular plugin type.

commands modified. Info filtering added to slurmctld.

loadable as shared aliases. Give slurm functions a prefix of "slurm_" to
insure that no user functions with the same name is used instead.
[For some reason a this same commit a minute ago did not seem to take effect.]

right away and treat failure as a fatal error.

write to.

    slurmctld startup
Create StateSaveLocation directory if changes via slurmctld reconfig

Properly support sinfo --state option value including "*" suffix.

  (which exhaust available memory).
Restructure slurmctld agent logic for higher throughput.

Fixes problem running very large number of simultaneous jobs when
their inactivity limit was reached due to a backlog of message.

data structure. This eliminates risks associated with re-reading slurm.conf.