<div dir="ltr">Ah, that&#39;s really weird. I&#39;m pretty sure that nothing ever made write changes to /export on either machine, so I wonder how the hard links ended up being split. I&#39;ll indeed clean up the .glusterfs directory and keep close tabs on Gluster&#39;s repair.<br><div><br></div><div>Glustershd.log and the client mount logs (data.log and gluster.log at least) on the client are empty and nothing appears when I read the mismatching studies.dat file.</div><div><br></div><div>Thanks for your help!</div><div>Sjors</div></div><br><div class="gmail_quote"><div dir="ltr">Op zo 7 jun. 2015 om 22:10 schreef Joe Julian &lt;<a href="mailto:joe@julianfamily.org">joe@julianfamily.org</a>&gt;:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
  
    
  
  <div bgcolor="#FFFFFF" text="#000000">
    (oops... I hate when I reply off-list)<br>
    <br>
    That warning should, imho, be an error. That&#39;s saying that the
    handle, which should be a hardlink to the file, doesn&#39;t have a
    matching inode. It should if it&#39;s a hardlink.<br>
    <br>
    If it were me, I would:<br>
    <br>
        find /export/sdb1/data/.glusterfs -type f -links 1 -print0 |
    xargs /bin/rm<br>
    <br>
    This would clean up any handles that are not hardlinked where they
    should be and will allow gluster to repair them.<br>
    <br>
    Btw, the self-heal errors would be in glustershd.log and/or the
    client mount log(s), not (usually) the brick logs.</div><div bgcolor="#FFFFFF" text="#000000"><br>
    <br>
    <div>On 06/07/2015 12:21 PM, Sjors Gielen
      wrote:<br>
    </div>
    </div><div bgcolor="#FFFFFF" text="#000000"><blockquote type="cite">
      <div dir="ltr">Oops! Accidentally ran the command as non-root on
        Curacao, that&#39;s why there was no output. The actual output is:<br>
        <div><br>
        </div>
        <div>curacao#<span style="line-height:1.5;font-size:13.1999998092651px"> getfattr
            -m . -d -e hex /export/sdb1/data/Case/21000355/studies.dat</span></div>
        <div><span style="line-height:1.5;font-size:13.1999998092651px">getfattr:
            Removing leading &#39;/&#39; from absolute path names</span></div>
        <div><span style="line-height:1.5;font-size:13.1999998092651px">#
            file: export/sdb1/data/Case/21000355/studies.dat</span></div>
        <div><span style="line-height:1.5;font-size:13.1999998092651px">trusted.afr.data-client-0=0x000000000000000000000000</span></div>
        <div><span style="line-height:1.5;font-size:13.1999998092651px">trusted.afr.data-client-1=0x000000000000000000000000</span></div>
        <div><span style="line-height:1.5;font-size:13.1999998092651px">trusted.afr.dirty=0x000000000000000000000000</span></div>
        <div><span style="line-height:1.5;font-size:13.1999998092651px">trusted.gfid=0xfb34574974cf4804b8b80789738c0f81</span></div>
        <div><span style="line-height:1.5;font-size:13.1999998092651px"><br>
          </span></div>
        <div><span style="line-height:1.5;font-size:13.1999998092651px">For
            reference, the output on bonaire:</span></div>
        <div><span style="line-height:1.5;font-size:13.1999998092651px"><br>
          </span></div>
        <div><span style="line-height:1.5;font-size:13.1999998092651px">bonaire# </span><span style="line-height:1.5;font-size:13.1999998092651px">getfattr
            -m . -d -e hex /export/sdb1/data/Case/21000355/studies.dat</span></div>
        <div><span style="font-size:13.1999998092651px;line-height:1.5">getfattr:
            Removing leading &#39;/&#39; from absolute path names</span></div>
        <div><span style="line-height:1.5;font-size:13.1999998092651px">#
            file: export/sdb1/data/Case/21000355/studies.dat</span></div>
        <div><span style="line-height:1.5;font-size:13.1999998092651px">trusted.gfid=0xfb34574974cf4804b8b80789738c0f81</span></div>
      </div>
      <br>
      <div class="gmail_quote">
        <div dir="ltr">Op zo 7 jun. 2015 om 21:13 schreef Sjors Gielen
          &lt;<a href="mailto:sjors@sjorsgielen.nl" target="_blank">sjors@sjorsgielen.nl</a>&gt;:<br>
        </div>
        <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
          <div dir="ltr">
            <div>I&#39;m reading about quorums, I haven&#39;t set up anything
              like that yet.</div>
            <div><br>
            </div>
            <div>(In reply to Joe Julian, who responded off-list)</div>
            <div><br>
            </div>
            The output of getfattr on bonaire:
            <div><br>
            </div>
            <div>bonaire# getfattr -m . -d -e hex
              /export/sdb1/data/Case/21000355/studies.dat</div>
            <div>getfattr: Removing leading &#39;/&#39; from absolute path names</div>
            <div># file: export/sdb1/data/Case/21000355/studies.dat</div>
            <div>trusted.gfid=0xfb34574974cf4804b8b80789738c0f81</div>
            <div><br>
            </div>
            <div>On curacao, the command gives no output.</div>
            <div><br>
            </div>
            <div>From `gluster volume status`, it seems that while the
              &quot;brick curacao:/export/sdb1/data&quot; is online, it has no
              associated port number. Curacao can connect to the port
              number provided by Bonaire just fine. There are no
              firewalls on/between the two machines, they are on the
              same subnet connected by Ethernet cables and two switches.</div>
            <div><br>
            </div>
            <div>By the way, warning messages just started appearing to
              /var/log/glusterfs/bricks/export-sdb1-data.log on Bonaire
              saying &quot;mismatching ino/dev between file X and handle Y&quot;,
              though, maybe only just now even though I started the full
              self-heal hours ago.</div>
            <div><br>
            </div>
            <div>[2015-06-07 19:10:39.624393] W
              [posix-handle.c:727:posix_handle_hard] 0-data-posix:
              mismatching ino/dev between file
              /export/sdb1/data/Archive/S21/21008971/studies.dat
              (9127104621/2065) and handle
              /export/sdb1/data/.glusterfs/97/c2/97c2a65d-36e0-4566-a5c1-5925f97af1fd
              (9190215976/2065)</div>
            <div><br>
            </div>
            <div>Thanks again!</div>
            <div>Sjors</div>
            <div><br>
              <div class="gmail_quote">
                <div dir="ltr">Op zo 7 jun. 2015 om 19:13 schreef Sjors
                  Gielen &lt;<a href="mailto:sjors@sjorsgielen.nl" target="_blank">sjors@sjorsgielen.nl</a>&gt;:<br>
                </div>
              </div>
            </div>
          </div>
          <div dir="ltr">
            <div>
              <div class="gmail_quote">
                <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
                  <div dir="ltr">Hi all,
                    <div><br>
                    </div>
                    <div>I work at a small, 8-person company that uses
                      Gluster for its primary data storage. We have a
                      volume called &quot;data&quot; that is replicated over two
                      servers (details below). This worked perfectly for
                      over a year, but lately we&#39;ve been noticing some
                      mismatches between the two bricks, so it seems
                      there has been some split-brain situation that is
                      not being detected or resolved. I have two
                      questions about this:</div>
                    <div><br>
                    </div>
                    <div>1) I expected Gluster to (eventually) detect a
                      situation like this; why doesn&#39;t it?</div>
                    <div>2) How do I fix this situation? I&#39;ve tried an
                      explicit &#39;heal&#39;, but that didn&#39;t seem to change
                      anything.</div>
                    <div><br>
                    </div>
                    <div>Thanks a lot for your help!</div>
                    <div>Sjors</div>
                    <div><br>
                    </div>
                    <div>------8&lt;------</div>
                    <div><br>
                    </div>
                    <div>Volume &amp; peer info: <a href="http://pastebin.com/PN7tRXdU" target="_blank">http://pastebin.com/PN7tRXdU</a></div>
                    <div>curacao# md5sum
                      /export/sdb1/data/Case/21000355/studies.dat</div>
                    <div>7bc2daec6be953ffae920d81fe6fa25c</div>
                    <div>/export/sdb1/data/Case/21000355/studies.dat<br>
                    </div>
                    <div>bonaire# md5sum
                      /export/sdb1/data/Case/21000355/studies.dat</div>
                    <div>28c950a1e2a5f33c53a725bf8cd72681
                      /export/sdb1/data/Case/21000355/studies.dat<br>
                    </div>
                    <div><br>
                    </div>
                    <div># mallorca is one of the clients</div>
                    <div>mallorca# md5sum
                      /data/Case/21000355/studies.dat</div>
                    <div>7bc2daec6be953ffae920d81fe6fa25c 
                      /data/Case/21000355/studies.dat</div>
                    <div><br>
                    </div>
                    <div>I expected an input/output error after reading
                      this file, because of the split-brain situation,
                      but got none. There are no entries in the
                      GlusterFS logs of either bonaire or curacao.</div>
                    <div><br>
                    </div>
                    <div>bonaire# gluster volume heal data full</div>
                    <div>Launching heal operation to perform full self
                      heal on volume data has been successful</div>
                    <div>Use heal info commands to check status</div>
                    <div>bonaire# gluster volume heal data info</div>
                    <div>Brick bonaire:/export/sdb1/data/</div>
                    <div>Number of entries: 0<br>
                    </div>
                    <div><br>
                    </div>
                    <div>Brick curacao:/export/sdb1/data/</div>
                    <div>Number of entries: 0</div>
                    <div><br>
                    </div>
                    <div>(Same output on curacao, and hours after this,
                      the md5sums on both bricks still differ.)</div>
                    <div><br>
                    </div>
                    <div>curacao# gluster --version</div>
                    <div>glusterfs 3.6.2 built on Mar  2 2015 14:05:34</div>
                    <div>Repository revision: git://<a href="http://git.gluster.com/glusterfs.git" target="_blank">git.gluster.com/glusterfs.git</a></div>
                    <div>(Same version on Bonaire)</div>
                  </div>
                </blockquote>
              </div>
            </div>
          </div>
        </blockquote>
      </div>
      <br>
      <fieldset></fieldset>
      <br>
      </blockquote></div><div bgcolor="#FFFFFF" text="#000000"><blockquote type="cite"><pre>_______________________________________________
Gluster-users mailing list
<a href="mailto:Gluster-users@gluster.org" target="_blank">Gluster-users@gluster.org</a>
<a href="http://www.gluster.org/mailman/listinfo/gluster-users" target="_blank">http://www.gluster.org/mailman/listinfo/gluster-users</a></pre>
    </blockquote></div>

_______________________________________________<br>
Gluster-users mailing list<br>
<a href="mailto:Gluster-users@gluster.org" target="_blank">Gluster-users@gluster.org</a><br>
<a href="http://www.gluster.org/mailman/listinfo/gluster-users" target="_blank">http://www.gluster.org/mailman/listinfo/gluster-users</a></blockquote></div>