<html>
  <head>
    <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    On 05/10/17 14:18, Pat Haley wrote:<br>
    <blockquote cite="mid:6dd8f0b6-016a-59f5-3fdb-afc9f1bb2f0f@mit.edu"
      type="cite">
      <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
      <br>
      Hi Pranith,<br>
      <br>
      Since we are mounting the partitions as the bricks, I tried the dd
      test writing to
      &lt;brick-path&gt;/.glusterfs/&lt;file-to-be-removed-after-test&gt;.
      The results without oflag=sync were 1.6 Gb/s (faster than gluster
      but not as fast as I was expecting given the 1.2 Gb/s to the
      no-gluster area w/ fewer disks).<br>
      <br>
      Pat<br>
      <br>
    </blockquote>
    <br>
    Is that true for every disk? If you're choosing the same filename
    every time for your dd test, you're likely only doing that test
    against one disk. If that disk is slow, you would get the same
    results every time despite other disks performing normally.<br>
    <br>
    <blockquote cite="mid:6dd8f0b6-016a-59f5-3fdb-afc9f1bb2f0f@mit.edu"
      type="cite"> <br>
      <div class="moz-cite-prefix">On 05/10/2017 01:27 PM, Pranith Kumar
        Karampuri wrote:<br>
      </div>
      <blockquote
cite="mid:CAOgeEnZjAG579+QtiGeViBsgJp0hNXx-x1J-4+1_4DuLU_PQgg@mail.gmail.com"
        type="cite">
        <meta http-equiv="Content-Type" content="text/html;
          charset=UTF-8">
        <div dir="ltr"><br>
          <div class="gmail_extra"><br>
            <div class="gmail_quote">On Wed, May 10, 2017 at 10:15 PM,
              Pat Haley <span dir="ltr">&lt;<a moz-do-not-send="true"
                  href="mailto:phaley@mit.edu" target="_blank">phaley@mit.edu</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"> <br>
                  Hi Pranith,<br>
                  <br>
                  Not entirely sure (this isn't my area of expertise). 
                  I'll run your answer by some other people who are more
                  familiar with this.<br>
                  <br>
                  I am also uncertain about how to interpret the results
                  when we also add the dd tests writing to the /home
                  area (no gluster, still on the same machine)<br>
                  <ul>
                    <li>dd test without oflag=sync (rough average of
                      multiple tests)<br>
                    </li>
                    <ul>
                      <li>gluster w/ fuse mount : 570 Mb/s</li>
                      <li>gluster w/ nfs mount:  390 Mb/s</li>
                      <li>nfs (no gluster):  1.2 Gb/s</li>
                    </ul>
                    <li>dd test with oflag=sync (rough average of
                      multiple tests)</li>
                    <ul>
                      <li>gluster w/ fuse mount:  5 Mb/s</li>
                      <li>gluster w/ nfs mount:  200 Mb/s</li>
                      <li>nfs (no gluster): 20 Mb/s<br>
                      </li>
                    </ul>
                  </ul>
                  Given that the non-gluster area is a RAID-6 of 4 disks
                  while each brick of the gluster area is a RAID-6 of 32
                  disks, I would naively expect the writes to the
                  gluster area to be roughly 8x faster than to the
                  non-gluster.<br>
                </div>
              </blockquote>
              <div><br>
              </div>
              <div>I think a better test is to try and write to a file
                using nfs without any gluster to a location that is not
                inside the brick but someother location that is on same
                disk(s). If you are mounting the partition as the brick,
                then we can write to a file inside .glusterfs directory,
                something like
                &lt;brick-path&gt;/.glusterfs/&lt;file-to-be-removed-after-test&gt;.
                <br>
                <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"> <br>
                  I still think we have a speed issue, I can't tell if
                  fuse vs nfs is part of the problem.</div>
              </blockquote>
              <div><br>
              </div>
              I got interested in the post because I read that fuse
              speed is lesser than nfs speed which is counter-intuitive
              to my understanding. So wanted clarifications. Now that I
              got my clarifications where fuse outperformed nfs without
              sync, we can resume testing as described above and try to
              find what it is. Based on your email-id I am guessing you
              are from Boston and I am from Bangalore so if you are okay
              with doing this debugging for multiple days because of
              timezones, I will be happy to help. Please be a bit
              patient with me, I am under a release crunch but I am very
              curious with the problem you posted.<br>
              <br>
            </div>
            <div class="gmail_quote">
              <blockquote class="gmail_quote" style="margin:0 0 0
                .8ex;border-left:1px #ccc solid;padding-left:1ex">
                <div bgcolor="#FFFFFF" text="#000000">  Was there
                  anything useful in the profiles?<span class="HOEnZb"><font
                      color="#888888"><br>
                    </font></span></div>
              </blockquote>
              <div><br>
              </div>
              <div>Unfortunately profiles didn't help me much, I think
                we are collecting the profiles from an active volume, so
                it has a lot of information that is not pertaining to dd
                so it is difficult to find the contributions of dd. So I
                went through your post again and found something I
                didn't pay much attention to earlier i.e. oflag=sync, so
                did my own tests on my setup with FUSE so sent that
                reply.<br>
                 <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"><span
                    class="HOEnZb"><font color="#888888"> <br>
                      Pat</font></span>
                  <div>
                    <div class="h5"><br>
                      <br>
                      <br>
                      <div class="m_-1330903101597946930moz-cite-prefix">On
                        05/10/2017 12:15 PM, Pranith Kumar Karampuri
                        wrote:<br>
                      </div>
                      <blockquote type="cite">
                        <div dir="ltr">
                          <div>
                            <div>Okay good. At least this validates my
                              doubts. Handling O_SYNC in gluster NFS and
                              fuse is a bit different.<br>
                            </div>
                            When application opens a file with O_SYNC on
                            fuse mount then each write syscall has to be
                            written to disk as part of the syscall where
                            as in case of NFS, there is no concept of
                            open. NFS performs write though a handle
                            saying it needs to be a synchronous write,
                            so write() syscall is performed first then
                            it performs fsync(). so an write on an fd
                            with O_SYNC becomes write+fsync. I am
                            suspecting that when multiple threads do
                            this write+fsync() operation on the same
                            file, multiple writes are batched together
                            to be written do disk so the throughput on
                            the disk is increasing is my guess.<br>
                            <br>
                          </div>
                          Does it answer your doubts?<br>
                        </div>
                        <div class="gmail_extra"><br>
                          <div class="gmail_quote">On Wed, May 10, 2017
                            at 9:35 PM, Pat Haley <span dir="ltr">&lt;<a
                                moz-do-not-send="true"
                                href="mailto:phaley@mit.edu"
                                target="_blank">phaley@mit.edu</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"> <br>
                                Without the oflag=sync and only a single
                                test of each, the FUSE is going faster
                                than NFS:<br>
                                <br>
                                FUSE:<br>
                                <tt>mseas-data2(dri_nascar)% dd
                                  if=/dev/zero count=4096 bs=1048576
                                  of=zeros.txt conv=sync</tt><tt><br>
                                </tt><tt>4096+0 records in</tt><tt><br>
                                </tt><tt>4096+0 records out</tt><tt><br>
                                </tt><tt>4294967296 bytes (4.3 GB)
                                  copied, 7.46961 s, 575 MB/s</tt><tt><br>
                                </tt><tt><br>
                                  <br>
                                </tt>NFS<br>
                                <tt>mseas-data2(HYCOM)% dd if=/dev/zero
                                  count=4096 bs=1048576 of=zeros.txt
                                  conv=sync</tt><tt><br>
                                </tt><tt>4096+0 records in</tt><tt><br>
                                </tt><tt>4096+0 records out</tt><tt><br>
                                </tt><tt>4294967296 bytes (4.3 GB)
                                  copied, 11.4264 s, 376 MB/s</tt>
                                <div>
                                  <div class="m_-1330903101597946930h5"><tt><br>
                                    </tt><tt><br>
                                    </tt><tt><br>
                                    </tt>
                                    <div
                                      class="m_-1330903101597946930m_7039630128565981365moz-cite-prefix">On
                                      05/10/2017 11:53 AM, Pranith Kumar
                                      Karampuri wrote:<br>
                                    </div>
                                    <blockquote type="cite">
                                      <div dir="ltr">Could you let me
                                        know the speed without
                                        oflag=sync on both the mounts?
                                        No need to collect profiles.<br>
                                      </div>
                                      <div class="gmail_extra"><br>
                                        <div class="gmail_quote">On Wed,
                                          May 10, 2017 at 9:17 PM, Pat
                                          Haley <span dir="ltr">&lt;<a
                                              moz-do-not-send="true"
                                              href="mailto:phaley@mit.edu"
                                              target="_blank">phaley@mit.edu</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"> <br>
                                              Here is what I see now:<br>
                                              <br>
                                              <tt>[root@mseas-data2 ~]#
                                                gluster volume info</tt><span><tt><br>
                                                </tt><tt> </tt><tt><br>
                                                </tt><tt>Volume Name:
                                                  data-volume</tt><tt><br>
                                                </tt><tt>Type:
                                                  Distribute</tt><tt><br>
                                                </tt><tt>Volume ID:
                                                  c162161e-2a2d-4dac-b015-f31fd8<wbr>9ceb18</tt><tt><br>
                                                </tt><tt>Status: Started</tt><tt><br>
                                                </tt><tt>Number of
                                                  Bricks: 2</tt><tt><br>
                                                </tt><tt>Transport-type:
                                                  tcp</tt><tt><br>
                                                </tt><tt>Bricks:</tt><tt><br>
                                                </tt><tt>Brick1:
                                                  mseas-data2:/mnt/brick1</tt><tt><br>
                                                </tt><tt>Brick2:
                                                  mseas-data2:/mnt/brick2</tt><tt><br>
                                                </tt><tt>Options
                                                  Reconfigured:</tt><tt><br>
                                                </tt></span><tt>diagnostics.count-fop-hits:
                                                on</tt><tt><br>
                                              </tt><tt>diagnostics.latency-measuremen<wbr>t:
                                                on</tt><tt><br>
                                              </tt><tt>nfs.exports-auth-enable:
                                                on</tt><tt><br>
                                              </tt><tt>diagnostics.brick-sys-log-leve<wbr>l:
                                                WARNING</tt><span><tt><br>
                                                </tt><tt>performance.readdir-ahead:
                                                  on</tt><tt><br>
                                                </tt><tt>nfs.disable: on</tt><tt><br>
                                                </tt><tt>nfs.export-volumes:
                                                  off</tt><tt><br>
                                                </tt><br>
                                                <br>
                                                <br>
                                              </span>
                                              <div>
                                                <div
                                                  class="m_-1330903101597946930m_7039630128565981365h5">
                                                  <div
class="m_-1330903101597946930m_7039630128565981365m_5079054141158038028moz-cite-prefix">On
                                                    05/10/2017 11:44 AM,
                                                    Pranith Kumar
                                                    Karampuri wrote:<br>
                                                  </div>
                                                  <blockquote
                                                    type="cite">
                                                    <div dir="ltr">
                                                      <div>Is this the
                                                        volume info you
                                                        have?<br>
                                                        <br>
                                                        <pre>&gt;<i>     [<a moz-do-not-send="true" href="http://www.gluster.org/mailman/listinfo/gluster-users" target="_blank">root at mseas-data2</a> ~]# gluster volume info
</i>&gt;<i>
</i>&gt;<i>     Volume Name: data-volume
</i>&gt;<i>     Type: Distribute
</i>&gt;<i>     Volume ID: c162161e-2a2d-4dac-b015-f31fd8<wbr>9ceb18
</i>&gt;<i>     Status: Started
</i>&gt;<i>     Number of Bricks: 2
</i>&gt;<i>     Transport-type: tcp
</i>&gt;<i>     Bricks:
</i>&gt;<i>     Brick1: mseas-data2:/mnt/brick1
</i>&gt;<i>     Brick2: mseas-data2:/mnt/brick2
</i>&gt;<i>     Options Reconfigured:
</i>&gt;<i>     performance.readdir-ahead: on
</i>&gt;<i>     nfs.disable: on
</i>&gt;<i>     nfs.export-volumes: off

</i></pre>
        </div>
        ​I copied this from old thread from 2016. This is distribute
        volume. Did you change any of the options in between?

      </div>
    </blockquote>
    

    </div></div><span><pre class="m_-1330903101597946930m_7039630128565981365m_5079054141158038028moz-signature" cols="72">-- 

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=<wbr>-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=<wbr>-=-=-
Pat Haley                          Email:  <a moz-do-not-send="true" class="m_-1330903101597946930m_7039630128565981365m_5079054141158038028moz-txt-link-abbreviated" href="mailto:phaley@mit.edu" target="_blank">phaley@mit.edu</a>
Center for Ocean Engineering       Phone:  (617) 253-6824
Dept. of Mechanical Engineering    Fax:    (617) 253-8125
MIT, Room 5-213                    <a moz-do-not-send="true" class="m_-1330903101597946930m_7039630128565981365m_5079054141158038028moz-txt-link-freetext" href="http://web.mit.edu/phaley/www/" target="_blank">http://web.mit.edu/phaley/www/</a>
77 Massachusetts Avenue
Cambridge, MA  02139-4301
</pre>
  </span></div>

</blockquote></div>


-- 
<div class="m_-1330903101597946930m_7039630128565981365gmail_signature" data-smartmail="gmail_signature"><div dir="ltr">Pranith
</div></div>
</div>



</blockquote>
<pre class="m_-1330903101597946930m_7039630128565981365moz-signature" cols="72">-- 

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=<wbr>-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=<wbr>-=-=-
Pat Haley                          Email:  <a moz-do-not-send="true" class="m_-1330903101597946930m_7039630128565981365moz-txt-link-abbreviated" href="mailto:phaley@mit.edu" target="_blank">phaley@mit.edu</a>
Center for Ocean Engineering       Phone:  (617) 253-6824
Dept. of Mechanical Engineering    Fax:    (617) 253-8125
MIT, Room 5-213                    <a moz-do-not-send="true" class="m_-1330903101597946930m_7039630128565981365moz-txt-link-freetext" href="http://web.mit.edu/phaley/www/" target="_blank">http://web.mit.edu/phaley/www/</a>
77 Massachusetts Avenue
Cambridge, MA  02139-4301
</pre></div></div></div></blockquote></div>


-- 
<div class="m_-1330903101597946930gmail_signature" data-smartmail="gmail_signature"><div dir="ltr">Pranith
</div></div>
</div>



</blockquote>
<pre class="m_-1330903101597946930moz-signature" cols="72">-- 

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=<wbr>-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=<wbr>-=-=-
Pat Haley                          Email:  <a moz-do-not-send="true" class="m_-1330903101597946930moz-txt-link-abbreviated" href="mailto:phaley@mit.edu" target="_blank">phaley@mit.edu</a>
Center for Ocean Engineering       Phone:  (617) 253-6824
Dept. of Mechanical Engineering    Fax:    (617) 253-8125
MIT, Room 5-213                    <a moz-do-not-send="true" class="m_-1330903101597946930moz-txt-link-freetext" href="http://web.mit.edu/phaley/www/" target="_blank">http://web.mit.edu/phaley/www/</a>
77 Massachusetts Avenue
Cambridge, MA  02139-4301
</pre></div></div></div></blockquote></div>


-- 
<div class="gmail_signature" data-smartmail="gmail_signature"><div dir="ltr">Pranith
</div></div>
</div></div>



</blockquote>
<pre class="moz-signature" cols="72">-- 

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Pat Haley                          Email:  <a moz-do-not-send="true" class="moz-txt-link-abbreviated" href="mailto:phaley@mit.edu">phaley@mit.edu</a>
Center for Ocean Engineering       Phone:  (617) 253-6824
Dept. of Mechanical Engineering    Fax:    (617) 253-8125
MIT, Room 5-213                    <a moz-do-not-send="true" class="moz-txt-link-freetext" href="http://web.mit.edu/phaley/www/">http://web.mit.edu/phaley/www/</a>
77 Massachusetts Avenue
Cambridge, MA  02139-4301
</pre>

<fieldset class="mimeAttachmentHeader"></fieldset>
<pre wrap="">_______________________________________________
Gluster-users mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Gluster-users@gluster.org">Gluster-users@gluster.org</a>
<a class="moz-txt-link-freetext" href="http://lists.gluster.org/mailman/listinfo/gluster-users">http://lists.gluster.org/mailman/listinfo/gluster-users</a></pre>

</blockquote>
</body></html>