<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  </head>
  <body>
    Angelos,<br>
    <br>
    I'm glad you mentioned UnkillableStepProgram.  We meant to look at
    that a while ago but forgot about it.  That will be very useful for
    us as well, though the answer for us is pretty much always Lustre
    problems.<br>
    <br>
    Ryan<br>
    <br>
    <div class="moz-cite-prefix">On 7/22/20 1:02 PM, Angelos Ching
      wrote:<br>
    </div>
    <blockquote type="cite"
      cite="mid:A6E5FB6B-F87C-4840-9981-8B5C5A4FEA01@clustertech.com">
      <meta http-equiv="content-type" content="text/html; charset=UTF-8">
      Agreed. You may also want to write a script that gather the list
      of program in "D state" (kernel wait) and print their stack; and
      configure it as UnkillableStepProgram so that you can capture the
      program and relevant system callS that caused the job to become
      unkillable / timed out exiting for further troubleshooting.
      <div><br>
        Regards,</div>
      <div>Angelos<br>
        <div dir="ltr">(Sent from mobile, please pardon me for typos and
          cursoriness.)</div>
        <div dir="ltr"><br>
          <blockquote type="cite">2020/07/23 0:41、Ryan Cox
            <a class="moz-txt-link-rfc2396E" href="mailto:ryan_cox@byu.edu"><ryan_cox@byu.edu></a>のメール:<br>
            <br>
          </blockquote>
        </div>
        <blockquote type="cite">
          <div dir="ltr">
            <meta http-equiv="Content-Type" content="text/html;
              charset=UTF-8">
            Ivan,<br>
            <br>
            Are you having I/O slowness? That is the most common cause
            for us. If it's not that, you'll want to look through all
            the reasons that it takes a long time for a process to
            actually die after a SIGKILL because one of those is the
            likely cause. Typically it's because the process is waiting
            for an I/O syscall to return. Sometimes swap death is the
            culprit, but usually not at the scale that you stated. 
            Maybe you could try reproducing the issue manually or
            putting something in epilog the see the state of the
            processes in the job's cgroup.<br>
            <br>
            Ryan<br>
            <br>
            <div class="moz-cite-prefix">On 7/22/20 10:24 AM, Ivan
              Kovanda wrote:<br>
            </div>
            <blockquote type="cite"
cite="mid:MWHPR11MB006111D9FE2985B00DFC220BF7790@MWHPR11MB0061.namprd11.prod.outlook.com">
              <meta http-equiv="Content-Type" content="text/html;
                charset=UTF-8">
              <meta name="Generator" content="Microsoft Word 15
                (filtered medium)">
              <style><!--
/* Font Definitions */
@font-face
        {font-family:"Cambria Math";
        panose-1:2 4 5 3 5 4 6 3 2 4;}
@font-face
        {font-family:Calibri;
        panose-1:2 15 5 2 2 2 4 3 2 4;}
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
        {margin:0in;
        margin-bottom:.0001pt;
        font-size:11.0pt;
        font-family:"Calibri",sans-serif;}
a:link, span.MsoHyperlink
        {mso-style-priority:99;
        color:#0563C1;
        text-decoration:underline;}
a:visited, span.MsoHyperlinkFollowed
        {mso-style-priority:99;
        color:#954F72;
        text-decoration:underline;}
span.EmailStyle17
        {mso-style-type:personal-compose;
        font-family:"Calibri",sans-serif;
        color:black;}
.MsoChpDefault
        {mso-style-type:export-only;
        font-family:"Calibri",sans-serif;}</style><!--[if gte mso 9]><xml>
<o:shapedefaults v:ext="edit" spidmax="1026" />
</xml><![endif]--><!--[if gte mso 9]><xml>
<o:shapelayout v:ext="edit">
<o:idmap v:ext="edit" data="1" />
</o:shapelayout></xml><![endif]-->
              <div class="WordSection1">
                <p class="MsoNormal"><span style="color:black">Dear
                    slurm community,<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black"><o:p> </o:p></span></p>
                <p class="MsoNormal"><span style="color:black">Currently
                    running slurm version 18.08.4<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black"><o:p> </o:p></span></p>
                <p class="MsoNormal"><span style="color:black">We have
                    been experiencing an issue causing any nodes a slurm
                    job was submitted to to "drain".<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black">From what
                    I've seen, it appears that there is a problem with
                    how slurm is cleaning up the job with the SIGKILL
                    process.<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black"><o:p> </o:p></span></p>
                <p class="MsoNormal"><span style="color:black">I've
                    found this slurm article (<a
                      class="moz-txt-link-freetext"
                      href="https://slurm.schedmd.com/troubleshoot.html#completing"
                      moz-do-not-send="true">https://slurm.schedmd.com/troubleshoot.html#completing</a>)
                    , which has a section titled "Jobs and nodes are
                    stuck in COMPLETING state", where it recommends
                    increasing the "UnkillableStepTimeout" in the
                    slurm.conf , but all that has done is prolong the
                    time it takes for the job to timeout. <o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black">The
                    default time for the "UnkillableStepTimeout" is 60
                    seconds.<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black"><o:p> </o:p></span></p>
                <p class="MsoNormal"><span style="color:black">After the
                    job completes, it stays in the CG (completing)
                    status for the 60 seconds, then the nodes the job
                    was submitted to go to drain status.<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black"><o:p> </o:p></span></p>
                <p class="MsoNormal"><span style="color:black">On the
                    headnode running slurmctld, I am seeing this in the
                    log - /var/log/slurmctld:<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black">--------------------------------------------------------------------------------------------------------------------------------------------<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black">[2020-07-21T22:40:03.000]
                    update_node: node node001 reason set to: Kill task
                    failed<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black">[2020-07-21T22:40:03.001]
                    update_node: node node001 state set to DRAINING<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black"><o:p> </o:p></span></p>
                <p class="MsoNormal"><span style="color:black">On the
                    compute node, I am seeing this in the log -
                    /var/log/slurmd<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black">--------------------------------------------------------------------------------------------------------------------------------------------<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black">[2020-07-21T22:38:33.110]
                    [1485.batch] done with job<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black">[2020-07-21T22:38:33.110]
                    [1485.extern] Sent signal 18 to 1485.4294967295<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black">[2020-07-21T22:38:33.111]
                    [1485.extern] Sent signal 15 to 1485.4294967295<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black">[2020-07-21T22:39:02.820]
                    [1485.extern] Sent SIGKILL signal to 1485.4294967295<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black">[2020-07-21T22:40:03.000]
                    [1485.extern] error: *** EXTERN STEP FOR 1485 STEPD
                    TERMINATED ON node001 AT 2020-07-21T22:40:02 DUE TO
                    JOB NOT ENDING WITH SIGNALS ***<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black"><o:p> </o:p></span></p>
                <p class="MsoNormal"><span style="color:black"><o:p> </o:p></span></p>
                <p class="MsoNormal"><span style="color:black">I've
                    tried restarting the SLURMD daemon on the compute
                    nodes, and even completing rebooting a few computes
                    nodes (node001, node002) . <o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black">From what
                    I've seen were experiencing this on all nodes in the
                    cluster. <o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black">I've yet
                    to restart the headnode because there are still
                    active jobs on the system so I don't want to
                    interrupt those.<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black"><o:p> </o:p></span></p>
                <p class="MsoNormal"><span style="color:black"><o:p> </o:p></span></p>
                <p class="MsoNormal"><span style="color:black">Thank you
                    for your time,<o:p></o:p></span></p>
                <p class="MsoNormal"><span style="color:black">Ivan<o:p></o:p></span></p>
                <p class="MsoNormal"><o:p> </o:p></p>
              </div>
            </blockquote>
          </div>
        </blockquote>
      </div>
    </blockquote>
    <br>
  </body>
</html>