<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <p><br>
    </p>
    <br>
    <div class="moz-cite-prefix">On 07/24/2018 02:56 PM, Raghavendra
      Gowdappa wrote:<br>
    </div>
    <blockquote type="cite"
cite="mid:CAFkORY8-FWYY433UJRrOtO4BOLDyHSEOVPar_a7RwC_RMxLzRA@mail.gmail.com">
      <div dir="ltr">
        <div>All,<br>
          <br>
        </div>
        I was trying to debug regression failures on [1] and observed
        that split-brain-resolution.t was failing consistently.<br>
        <div>
          <div>
            <div><br>
              =========================<br>
              TEST 45 (line 88): 0 get_pending_heal_count patchy<br>
              ./tests/basic/afr/split-brain-resolution.t .. 45/45 RESULT
              45: 1<br>
              ./tests/basic/afr/split-brain-resolution.t .. Failed 17/45
              subtests <br>
              <br>
              Test Summary Report<br>
              -------------------<br>
              ./tests/basic/afr/split-brain-resolution.t (Wstat: 0
              Tests: 45 Failed: 17)<br>
                Failed tests:  24-26, 28-36, 41-45<br>
            </div>
            <div><br>
            </div>
            <div><br>
            </div>
            <div>On probing deeper, I observed a curious fact - on most
              of the failures stat was not served from md-cache, but
              instead was wound down to afr which failed stat with EIO
              as the file was in split brain. So, I did another test:</div>
            <div>* disabled md-cache</div>
            <div>* mount glusterfs with attribute-timeout 0 and
              entry-timeout 0</div>
            <div><br>
            </div>
            <div>Now the test fails always. So, I think the test relied
              on stat requests being absorbed either by kernel attribute
              cache or md-cache. When its not happening stats are
              reaching afr and resulting in failures of cmds like
              getfattr etc. </div>
          </div>
        </div>
      </div>
    </blockquote>
    <br>
    This indeed seems to be the case.  Is there any way we can avoid the
    stat? When a getfattr is performed on the mount, aren't lookup +
    getfattr are the only fops that need to be hit in gluster? <br>
    -Ravi<br>
    <br>
    <blockquote type="cite"
cite="mid:CAFkORY8-FWYY433UJRrOtO4BOLDyHSEOVPar_a7RwC_RMxLzRA@mail.gmail.com">
      <div dir="ltr">
        <div>
          <div>
            <div>Thoughts?<br>
            </div>
            <div><br>
            </div>
            <div>[1] <a href="https://review.gluster.org/#/c/20549/"
                moz-do-not-send="true">https://review.gluster.org/#/c/20549/</a><br>
            </div>
          </div>
        </div>
      </div>
      <br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <br>
      <pre wrap="">_______________________________________________
Gluster-devel mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Gluster-devel@gluster.org">Gluster-devel@gluster.org</a>
<a class="moz-txt-link-freetext" href="https://lists.gluster.org/mailman/listinfo/gluster-devel">https://lists.gluster.org/mailman/listinfo/gluster-devel</a></pre>
    </blockquote>
    <br>
  </body>
</html>