<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <div class="moz-cite-prefix">On 08/23/12 01:18, Sunil Mushran wrote:<br>
    </div>
    <blockquote
cite="mid:CAEeiSHWQMVEPM1113ePCajYFGt-BS84gmLn=03v-_CvUQJ-owQ@mail.gmail.com"
      type="cite">Yes. WRITE_SYNC should be good. Not FUA.
      <div><br>
      </div>
      <div>Also, you may want to look into using io priorities. The code
        is all there. Just needs activation.<br>
      </div>
    </blockquote>
    Yes, I'll search the list to find them out.<br>
    <br>
    Thanks,<br>
    -Jeff<br>
    <blockquote
cite="mid:CAEeiSHWQMVEPM1113ePCajYFGt-BS84gmLn=03v-_CvUQJ-owQ@mail.gmail.com"
      type="cite">
      <div><br>
        <div class="gmail_quote">On Wed, Aug 22, 2012 at 10:13 AM,
          srinivas eeda <span dir="ltr">&lt;<a moz-do-not-send="true"
              href="mailto:srinivas.eeda@oracle.com" target="_blank">srinivas.eeda@oracle.com</a>&gt;</span>
          wrote:<br>
          <blockquote class="gmail_quote" style="margin:0 0 0
            .8ex;border-left:1px #ccc solid;padding-left:1ex">
            <div bgcolor="#FFFFFF" text="#000000">
              <div class="im"> <br>
                On 8/22/2012 7:17 AM, Jie Liu wrote:
                <blockquote type="cite"> Hi All,<br>
                  <br>
                  These days, I am investigating an issue regarding
                  OCFS2 unexpected reboot in some real world use cases.<br>
                  This problem occurred when the network status goes
                  south, when the disk IO load is too high, etc...<br>
                  I suspect it might caused by ocfs2 fencing if it's BIO
                  reading/writing can not be scheduled and processed
                  quickly, or<br>
                  something like this happened in the network IO
                  heartbeat thread.<br>
                  <br>
                  Now am trying to reproduce this problem locally.&nbsp; In
                  the meantime, I'd like to ping you guys with some
                  rough ideas<br>
                  to improve the disk IO heartbeat to see if they are
                  sounds reasonable or not.<br>
                  <br>
                  Firstly, if an OCFS2 node is suffer from heavy disk
                  IO, how about to fix the bio read/write to make this
                  IO request can not<br>
                  be preempted by other requests? e.g,&nbsp; for <span
                    style="margin-right:0"><span></span></span>o2hb_issue_node_write(),


                  currently, it do bio submission with WRITE only,<br>
                  'submit_bio(WRITE, bio)'. &nbsp; If we change the flag to
                  WRITE_SYNC, or even submit the request combine with
                  REQ_FUA,<br>
                  maybe could get highest priority for disk IO request.<br>
                </blockquote>
              </div>
              This was submitted before by Noboru Iwamatsu and acked by
              sunil and tao but some how didn't get merged<br>
              <br>
              <a moz-do-not-send="true"
href="https://oss.oracle.com/pipermail/ocfs2-devel/2011-December/008438.html"
                target="_blank">https://oss.oracle.com/pipermail/ocfs2-devel/2011-December/008438.html</a><br>
              <blockquote type="cite">
                <div class="im"> <br>
                  Secondly, the comments for bio allocation at
                  o2hb_setup_one_bio() indicates that we can
                  pre-allocate bio instead of<br>
                  acquire for each time.&nbsp; But I have not saw any code
                  snippet doing such things in kernel. :(<br>
                  how about creating a private bio set for each
                  o2hb_region, so that we can do allocation out of it? <br>
                  maybe it's faster than do allocation from global bio
                  sets.&nbsp; Also, does it make sense if creating a memory
                  pool<br>
                  on each o2hb_region, so that we can have continuous
                  pages bind to those bios?<br>
                  <br>
                  <br>
                  Any comments are appreciated!<br>
                  <br>
                  Thanks,<br>
                  -Jeff<br>
                  <br>
                  <br>
                  <br>
                  <br>
                  <br>
                  <br>
                  <br>
                  <fieldset></fieldset>
                  <br>
                </div>
                <pre>_______________________________________________
Ocfs2-devel mailing list
<a moz-do-not-send="true" href="mailto:Ocfs2-devel@oss.oracle.com" target="_blank">Ocfs2-devel@oss.oracle.com</a>
<a moz-do-not-send="true" href="https://oss.oracle.com/mailman/listinfo/ocfs2-devel" target="_blank">https://oss.oracle.com/mailman/listinfo/ocfs2-devel</a></pre>
              </blockquote>
            </div>
            <br>
            _______________________________________________<br>
            Ocfs2-devel mailing list<br>
            <a moz-do-not-send="true"
              href="mailto:Ocfs2-devel@oss.oracle.com">Ocfs2-devel@oss.oracle.com</a><br>
            <a moz-do-not-send="true"
              href="https://oss.oracle.com/mailman/listinfo/ocfs2-devel"
              target="_blank">https://oss.oracle.com/mailman/listinfo/ocfs2-devel</a><br>
          </blockquote>
        </div>
        <br>
      </div>
    </blockquote>
    <br>
  </body>
</html>