<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  </head>
  <body>
    <div class="moz-text-flowed" style="font-family: -moz-fixed;
      font-size: 13px;" lang="x-unicode">Hi, <br>
      <br>
      The slurmctld.log shows (for this node): <br>
      <br>
      ... <br>
      <br>
      [2021-05-25T00:12:27.481] sched: Allocate JobId=3402729
      NodeList=gpu-t4-4x-ondemand-44 #CPUs=1
      Partition=gpu-t4-4x-ondemand <br>
      [2021-05-25T00:12:27.482] sched: Allocate JobId=3402730
      NodeList=gpu-t4-4x-ondemand-44 #CPUs=1
      Partition=gpu-t4-4x-ondemand <br>
      [2021-05-25T00:12:27.482] sched: Allocate JobId=3402731
      NodeList=gpu-t4-4x-ondemand-44 #CPUs=1
      Partition=gpu-t4-4x-ondemand <br>
      [2021-05-25T00:15:29.043] error: Nodes gpu-t4-4x-ondemand-44 not
      responding <br>
      <br>
      <br>
      sinfo -R initially doesn't show it as problematic. Though I see it
      goes into: <br>
      <br>
      gpu-t4-4x-ondemand                 up infinite      1  comp*
      gpu-t4-4x-ondemand-44 <br>
      <br>
      However, the node where slurmctld is running knows about it: <br>
      <br>
       host gpu-t4-4x-ondemand-44 <br>
      gpu-t4-4x-ondemand-44.virtual-cluster.local has address 10.4.64.11
      <br>
      <br>
      and I can log in to the node: <br>
      <br>
      # systemctl status slurmd <br>
      \u25cf slurmd.service - Slurm node daemon <br>
         Loaded: loaded (/usr/lib/systemd/system/slurmd.service;
      disabled; vendor preset: disabled) <br>
         Active: active (running) since Tue 2021-05-25 00:12:24 UTC; 48s
      ago <br>
        Process: 1874 ExecStart=/opt/slurm/sbin/slurmd $SLURMD_OPTIONS
      (code=exited, status=0/SUCCESS) <br>
       Main PID: 1876 (slurmd) <br>
          Tasks: 1 <br>
         Memory: 11.6M <br>
         CGroup: /system.slice/slurmd.service <br>
                 \u2514\u25001876 /opt/slurm/sbin/slurmd -f
      /etc/slurm/slurm.conf <br>
      <br>
      May 25 00:12:23 gpu-t4-4x-ondemand-44.virtual-cluster.local
      systemd[1]: Starting Slurm node daemon... <br>
      May 25 00:12:24 gpu-t4-4x-ondemand-44.virtual-cluster.local
      systemd[1]: Can't open PID file /var/run/slurmd.pid (yet?)...ory <br>
      May 25 00:12:24 gpu-t4-4x-ondemand-44.virtual-cluster.local
      systemd[1]: Started Slurm node daemon. <br>
      Hint: Some lines were ellipsized, use -l to show in full. <br>
      <br>
      later: <br>
      <br>
      sinfo: <br>
      <br>
      gpu-t4-4x-ondemand                 up infinite      1  idle*
      gpu-t4-4x-ondemand-44 <br>
      <br>
      root@service(eigen2):log# sinfo -R <br>
      REASON               USER      TIMESTAMP           NODELIST <br>
      Not responding       slurm     2021-05-25T00:45:40
      gpu-t4-4x-ondemand-44 <br>
      <br>
      and slurmctld.log: <br>
      <br>
      [2021-05-25T00:15:29.043] error: Nodes gpu-t4-4x-ondemand-44 not
      responding <br>
      [2021-05-25T00:19:16.397] error: Nodes gpu-t4-4x-ondemand-44 not
      responding, setting DOWN <br>
      [2021-05-25T00:20:02.092] powering down node gpu-t4-4x-ondemand-44
      <br>
      [2021-05-25T00:20:08.438] error: Nodes gpu-t4-4x-ondemand-44 not
      responding, setting DOWN <br>
      [2021-05-25T00:25:02.931] powering down node gpu-t4-4x-ondemand-44
      <br>
      [2021-05-25T00:25:04.903] error: Nodes gpu-t4-4x-ondemand-44 not
      responding, setting DOWN <br>
      [2021-05-25T00:30:01.247] powering down node gpu-t4-4x-ondemand-44
      <br>
      [2021-05-25T00:31:21.479] error: Nodes gpu-t4-4x-ondemand-44 not
      responding, setting DOWN <br>
      [2021-05-25T00:35:01.359] powering down node gpu-t4-4x-ondemand-44
      <br>
      [2021-05-25T00:35:41.756] error: Nodes gpu-t4-4x-ondemand-44 not
      responding, setting DOWN <br>
      [2021-05-25T00:40:01.671] powering down node gpu-t4-4x-ondemand-44
      <br>
      [2021-05-25T00:40:41.225] error: Nodes gpu-t4-4x-ondemand-44 not
      responding, setting DOWN <br>
      [2021-05-25T00:45:01.430] powering down node gpu-t4-4x-ondemand-44
      <br>
      [2021-05-25T00:45:40.071] error: Nodes gpu-t4-4x-ondemand-44 not
      responding, setting DOWN <br>
    </div>
    <p><br>
      This makes sense given what it thinks the state is.  However, it's
      unclear why it thinks it's non-responding given that slurmd is
      running and that it can be logged into.  <br>
    </p>
    <p> Herc</p>
    <p><br>
    </p>
  </body>
</html>