<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#ffffff" text="#000000">
    It shouldn't. But I have been known to be wrong before.<br>
    And testing with different commit values is easy enough.<br>
    <br>
    On 02/02/2011 06:03 PM, Henrique Fernandes wrote:
    <blockquote
      cite="mid:AANLkTimbM1M=jcC5qkg6yJh3ouCq1MHfCK4ZLXa6V5ut@mail.gmail.com"
      type="cite">Thanks for the tips.<br>
      <br>
      So you are saying that if both nodes have same load, inscrease
      comit time will not make any diferences ?<br>
      <br>
      <br>
      <br clear="all">
      []'sf.rique <br>
      <br>
      <br>
      <div class="gmail_quote">On Wed, Feb 2, 2011 at 11:54 PM, Sunil
        Mushran <span dir="ltr">&lt;<a moz-do-not-send="true"
            href="mailto:sunil.mushran@oracle.com">sunil.mushran@oracle.com</a>&gt;</span>
        wrote:<br>
        <blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt
          0.8ex; border-left: 1px solid rgb(204, 204, 204);
          padding-left: 1ex;">
          <div bgcolor="#ffffff" text="#000000"> # grep "Resources"
            /sys/kernel/debug/o2dlm/UUID/dlm_state<br>
            Lock Resources: 20713 (1139459)<br>
            <br>
            The first number is number of live resources. Second the
            number<br>
            of resources created for the life of the domain.<br>
            <br>
            If you see the the number of resources levelling off but the
            second<br>
            number incrementing, that _could_ mean that you could
            benefit with<br>
            a larger cache. But you have to supplement that number with
            the<br>
            workload. Say you are deleting and creating files. In that
            case, you<br>
            will see the same number movement but not because of the
            cache limit.<br>
            <br>
            We are in the process of adding more counters to provide
            more visibility<br>
            to the users.<br>
            <br>
            Well, using any software without support has its risks.<br>
            <br>
            Using noatime will help. Increasing commit will help if
            workload if largely<br>
            from one node only.
            <div>
              <div class="h5"><br>
                <br>
                On 02/02/2011 05:10 PM, Henrique Fernandes wrote:
                <blockquote type="cite">centos 5.5&nbsp; <br>
                  ocfs2 1.4<br>
                  The mail solution, dovecot+postifx+mailscaner&nbsp; uses
                  about 1 gb the machine has 1.5 in my tests i am rising
                  it to 3gb so it have 2gb of cached<br>
                  <br>
                  How many memory do you thing i should separeted for
                  cache ?<br>
                  <br>
                  When we made some tests&nbsp; ocfs2 1.4 had better
                  performance than 1.6 but or tests were very simple, an
                  script that writes lots of files and anothe rone that
                  reads it.<br>
                  <br>
                  And it is a problem to have ocfs2 1.6 in ths centos!!<br>
                  <br>
                  Should i use ocfs2 in production ?<br>
                  <br>
                  how about the commit and noatime configs ??<br>
                  <br>
                  <br>
                  thanks!!<br>
                  <br clear="all">
                  []'sf.rique <br>
                  <br>
                  <br>
                  <div class="gmail_quote">On Wed, Feb 2, 2011 at 9:27
                    PM, Sunil Mushran <span dir="ltr">&lt;<a
                        moz-do-not-send="true"
                        href="mailto:sunil.mushran@oracle.com"
                        target="_blank">sunil.mushran@oracle.com</a>&gt;</span>
                    wrote:<br>
                    <blockquote class="gmail_quote" style="margin: 0pt
                      0pt 0pt 0.8ex; border-left: 1px solid rgb(204,
                      204, 204); padding-left: 1ex;">
                      <div bgcolor="#ffffff" text="#000000"> version?
                        distro?<br>
                        <br>
                        This workload will benefit a lot with the
                        indexed directories available in<br>
                        ocfs2 1.6 (and mainline and sles11).<br>
                        <br>
                        The other thing to check is the amount memory in
                        the virtual machines.<br>
                        File systems need memory to cache the inodes. If
                        memory is lacking,<br>
                        the inodes are freed and have to be re-read from
                        disk time and again.<br>
                        While this is a problem even in a local fs, it
                        is a bigger problem in a cfs<br>
                        as a cfs needs to do lock mastery for the same
                        inode time and again.
                        <div>
                          <div><br>
                            <br>
                            On 02/02/2011 01:09 PM, Henrique Fernandes
                            wrote: </div>
                        </div>
                        <blockquote type="cite">
                          <div>
                            <div>Hello,<br>
                              <br>
                              First of all, i am new at the list and i
                              have several questions about ocfs2
                              performance.<br>
                              <br>
                              Where i am working i am having huge
                              performance problens with ocfs2.<br>
                              <br>
                              Let me tell my envoriment.<br>
                              <br>
                              3 Xen VirtualMachines withs ocfs2 mounting
                              an LUN exported over iSCSI. ( acctualy 3
                              LUNS, 3 ocfs2 clusters )<br>
                              <br>
                              I am not the one who configured the
                              envoriment, but it is making the
                              performance of my MAIL system to bad.<br>
                              <br>
                              Have about 9k accounts but only 4k are
                              active. It is a maildir system. ( postfix
                              + dovecot )<br>
                              <br>
                              Now that this performance problens are
                              afecting my system i am gonna try help to
                              tunning the ocfs2.<br>
                              <br>
                              Pretty much all default settings.<br>
                              <br>
                              OCFS2 is configured to write with ordered
                              mode. We know that changing to writeback
                              will make performance much better, but we
                              are not considering lose anydata, so it i
                              snot an option.<br>
                              <br>
                              Now we are going to implemente noatime
                              options in mount. This make better
                              performace ?<br>
                              <br>
                              Other one, how about the commit mount
                              options ? The default is set to 5s&nbsp; if i
                              increse it how is the potential data loss
                              in case we lost lose power?<br>
                              <br>
                              Does anyone have any other paramenter that
                              should help us ?<br>
                              <br>
                              Another info, the inscremental backup is
                              taking 10 to 12 hours. <br>
                              <br>
                              All nodes have VERY high I/O wait.<br>
                              <br>
                              Thanks to all!!<br>
                              <br>
                              If you could tell me any doc that i sould
                              read would be nice to!<br>
                              <br>
                              <br>
                              <br>
                              <br>
                              <br>
                              <br clear="all">
                              []'sf.rique <br>
                            </div>
                          </div>
                          <pre><fieldset></fieldset>
_______________________________________________
Ocfs2-users mailing list
<a moz-do-not-send="true" href="mailto:Ocfs2-users@oss.oracle.com" target="_blank">Ocfs2-users@oss.oracle.com</a>
<a moz-do-not-send="true" href="http://oss.oracle.com/mailman/listinfo/ocfs2-users" target="_blank">http://oss.oracle.com/mailman/listinfo/ocfs2-users</a></pre>
                        </blockquote>
                        <br>
                      </div>
                    </blockquote>
                  </div>
                  <br>
                </blockquote>
                <br>
              </div>
            </div>
          </div>
        </blockquote>
      </div>
      <br>
    </blockquote>
    <br>
  </body>
</html>