<html>
  <head>

    <meta http-equiv="content-type" content="text/html; charset=UTF-8">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <p>All,</p>
    <p>I have several clusters that are all connected to a standalone
      slurmdb server. They are not federated.</p>
    <p>I can check the queues and do everything from that system using
      -M <cluster> for most commands. However, if I try to get a
      shell within a job (eg: srun -M cluster-a --pty bash), it queues
      up, but when it tries to run, I get an error:</p>
    <div class="moz-text-html" lang="x-western"><font size="2"
        face="Calibri"><span style="font-size:11pt;">
          <div> </div>
        </span></font><font size="2" face="Calibri"><span
          style="font-size:11pt;">
          <div>$ srun -M clustera -n16 --pty bash</div>
          <div>srun: job 6 has been allocated resources</div>
          <div>srun: error: Error connecting, bad data: family = 2, port
            = 0</div>
          <div>srun: error: Task launch for 6.0 failed on node
            ip-0A312014: Communication connection failure</div>
          <div>srun: error: Application launch failed: Communication
            connection failure</div>
          <div>srun: Job step aborted: Waiting up to 32 seconds for job
            step to finish.</div>
          <div>srun: error: Timed out waiting for job step to complete</div>
          <div><br>
          </div>
          <div> </div>
          <div> </div>
        </span></font>And in the slurmctld log for the cluster master:</div>
    <div class="moz-text-html" lang="x-western"><span
        style="font-size:11pt;">
        <div>Dec 17 19:50:04 nastran-master slurmctld[54739]: error:
          slurm_receive_msg [10.49.32.20:44022]: Zero Bytes were
          transmitted or received</div>
        <div>Dec 17 19:50:07 nastran-master slurmctld[54739]: error:
          slurm_receive_msg [10.49.32.20:44046]: Zero Bytes were
          transmitted or received</div>
        <div>Dec 17 19:50:08 nastran-master slurmctld[54739]:
          update_node: node ip-0A312014 state set to DOWN</div>
        <div>Dec 17 19:50:08 nastran-master slurmctld[54739]: Node
          ip-0A312014 now responding</div>
        <div>Dec 17 19:50:08 nastran-master slurmctld[54739]: node
          ip-0A312014 returned to service</div>
        <div>Dec 17 19:50:09 nastran-master slurmctld[54739]: sched:
          Allocate JobId=6 NodeList=ip-0A312014 #CPUs=16 Partition=debug</div>
        <div>Dec 17 19:50:09 nastran-master slurmctld[54739]:
          job_step_signal JobId=6 StepId=0 not found</div>
        <div>Dec 17 19:50:41 nastran-master slurmctld[54739]:
          job_step_signal JobId=6 StepId=0 not found</div>
        <div>Dec 17 19:50:41 nastran-master slurmctld[54739]:
          _job_complete: JobId=6 WTERMSIG 105</div>
        <div>Dec 17 19:50:41 nastran-master slurmctld[54739]:
          _job_complete: JobId=6 done</div>
      </span></div>
    <div class="moz-text-html" lang="x-western"><br>
    </div>
    <p>Is this something that cannot be done from a system that is
      outside a federated cluster?</p>
    <p>Brian Andrus<br>
    </p>
  </body>
</html>