to the job).

improved failure recovery.

(e.g. "default"=="DEFAULT", "Yes"=="YES", etc.)

Nodes actually become usable when slurmd gets started or slurmctld
pings the node and it responds.

Remove non-agent based credential revoke.
Minor clean-up of code to handle non-responsive node.

state change to NOT_RESPOND or DOWN.

slurmctld now will ping slurmd periodically and flag non-responsive nodes as such.
slurmd now responds to ping RPC.

slurmctld now will ping slurmd periodically and flag non-responsive nodes as such.

Expanded error recovery logic on slurmd registration.
slurmd to register active job_id plus step_id on initial registration.

Expanded error recovery logic on slurmd registration.

   still not perfect

Terminate job if slurmd restart without it.

to slurmctld on restart.

Get default TmpFS from #define instead of hard-wire to "/tmp".

   more than 1 process per node (and generating random hw contexts)

   still needs work -- should we continue to use the socket
   or close it?

enhanced debugging added for agent, now passing node name along with address.

 o close fd on POLLERR in poll() loop

DejaGnu directory files).