<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    Adding Eli.<br>
    <br>
    <br>
    <div class="moz-cite-prefix">On 01/27/2014 02:50 PM, Andrew Lau
      wrote:<br>
    </div>
    <blockquote
cite="mid:CAD7dF9fNhvSsd+Oj2s+rJo4oSkwnZiU6H23tbY0XL12_JShfsw@mail.gmail.com"
      type="cite">
      <div dir="ltr">
        <div class="gmail_default" style="font-family:tahoma,sans-serif">Hi,</div>
        <div class="gmail_default" style="font-family:tahoma,sans-serif"><br>
        </div>
        <div class="gmail_default" style="font-family:tahoma,sans-serif">
          I think he was asking what if the power management device
          reported that the host was powered off. Then VMs should be
          brought back up as being off would essentially be the same as
          running a power cycle/reboot?</div>
        <div class="gmail_default" style="font-family:tahoma,sans-serif">
          <br>
        </div>
        <div class="gmail_default" style="font-family:tahoma,sans-serif">Another
          example I'm seeing is what happens if the whole host loses
          power and it's power management device then becomes
          unavailable (ie. not reachable) then you're stuck in the case
          where it requires manual intervention.&nbsp;</div>
        <div class="gmail_default" style="font-family:tahoma,sans-serif"><br>
        </div>
        <div class="gmail_default" style="font-family:tahoma,sans-serif">I
          would be interested to potentially see something like a
          timeout on those problematic VMs (eg. if nothing was read or
          write after x amount of time) then you could consider the host
          as offline? I guess then that adds a lot of risk..</div>
        <div class="gmail_extra">
          <br>
          <br>
          <div class="gmail_quote">On Mon, Jan 27, 2014 at 11:43 PM,
            Tareq Alayan <span dir="ltr">&lt;<a moz-do-not-send="true"
                href="mailto:talayan@redhat.com" target="_blank">talayan@redhat.com</a>&gt;</span>
            wrote:<br>
            <blockquote class="gmail_quote" style="margin:0 0 0
              .8ex;border-left:1px #ccc solid;padding-left:1ex">
              Hi,<br>
              <br>
              Power management makes use of special *dedicated* hardware
              in order to restart hosts independently of host OS. The
              engine connects to a power management devices using a
              *dedicated* network IP address.<br>
              The engine is capable of rebooting hosts that have entered
              a non-operational or non-responsive state,<br>
              The abilities provided by all power management devices
              are: check status, start, stop and recycle (restart)...<br>
              <br>
              In the case of non-responsive host: all of the VMs that
              are currently running on that host can also become
              non-responsive. However, the non-responsive host keeps
              locking the VM hard disk for all VMs it is running.
              Attempting to start a VM on a different host and assign
              the second host write privileges for the virtual machine
              hard disk image can cause data corruption.<br>
              Rebooting allows the engine to assume that the lock on a
              VM hard disk image has been released.<br>
              The engine can know for sure that the problematic host has
              been rebooted via the power management device and then it
              can start a VM from the problematic host on another host
              without risking data corruption.<br>
              Important note: A virtual machine that has been marked
              highly-available can not be safely started on a different
              host without the certainty that doing so will not cause
              data corruption.<br>
              <br>
              N-joy,<br>
              <br>
              --Tareq
              <div class="HOEnZb">
                <div class="h5"><br>
                  <br>
                  <br>
                  <br>
                  On 01/27/2014 02:05 PM, Dafna Ron wrote:<br>
                  <blockquote class="gmail_quote" style="margin:0 0 0
                    .8ex;border-left:1px #ccc solid;padding-left:1ex">
                    I am adding Tareq for the Power Management
                    implementation.<br>
                    <br>
                    Dafna<br>
                    <br>
                    <br>
                    On 01/27/2014 11:48 AM, Karli Sj&ouml;berg wrote:<br>
                    <blockquote class="gmail_quote" style="margin:0 0 0
                      .8ex;border-left:1px #ccc solid;padding-left:1ex">
                      On Mon, 2014-01-27 at 11:11 +0000, Dafna Ron
                      wrote:<br>
                      <blockquote class="gmail_quote" style="margin:0 0
                        0 .8ex;border-left:1px #ccc
                        solid;padding-left:1ex">
                        Powering off the host will never trigger vm
                        migration.<br>
                        As far as engine is concerned it just lost
                        connection to the host, but<br>
                        has no way of telling if the host is down or if
                        a router is down.<br>
                      </blockquote>
                      Can&acute;t it at least check with power management if
                      the Host status is down<br>
                      first?<br>
                      <br>
                      I mean, if the network is down there will be no
                      response from either PM<br>
                      or Host. But if PM is up and can tell you that the
                      Host is down, sounds<br>
                      rather clear cut to me...<br>
                      <br>
                      Seems to me the VM's would be restarted sooner if
                      the flow was altered<br>
                      to first check with PM if it&acute;s a network or Host
                      issue, and if Host<br>
                      issue, immediately restart VM's on another Host,
                      instead of waiting for<br>
                      a potentially problematic Host to boot up
                      eventually.<br>
                      <br>
                      /K<br>
                      <br>
                      <blockquote class="gmail_quote" style="margin:0 0
                        0 .8ex;border-left:1px #ccc
                        solid;padding-left:1ex">
                        since vm's can continue running on the host even
                        if engine has no access<br>
                        to it, starting the vm's on the second host can
                        cause split brain and<br>
                        data corruption.<br>
                        <br>
                        The way that the engine knows what's going on is
                        by sending heath check<br>
                        queries to the vdsm.<br>
                        Power management will try to reboot a host when
                        the health checks to<br>
                        vdsm will not be answered.<br>
                        So... if engine gets no reply and has no way of
                        rebooting the host, the<br>
                        host status will be changed to Non-Responsive
                        and the vm's will be<br>
                        unknown because engine has no way of knowing
                        what's happening with the<br>
                        vm's.<br>
                        Since reboot of the host will kill the vm's
                        running on it - this will<br>
                        never cause any vm migration but... along with
                        the High-Availability vm<br>
                        feature, you will be able to have some of the
                        vm's re-started on the<br>
                        second host after the host reboot (and that is
                        only if Power Management<br>
                        was confirmed as successful).<br>
                        <br>
                        VM migration is only triggered when:<br>
                        1. Cluster configuration states that the vm
                        should be migrated in case<br>
                        of failure<br>
                        2. Engine has access to the host - so the
                        failure is on the storage side<br>
                        and not the host side.<br>
                        3. the vms are not actively writing (although
                        there might be a new RFE<br>
                        for it).<br>
                        <br>
                        hope this clears things up<br>
                        <br>
                        Dafna<br>
                        <br>
                        <br>
                        <br>
                        On 01/27/2014 10:11 AM, Andrew Lau wrote:<br>
                        <blockquote class="gmail_quote" style="margin:0
                          0 0 .8ex;border-left:1px #ccc
                          solid;padding-left:1ex">
                          Hi,<br>
                          <br>
                          Have you got power management enabled?<br>
                          <br>
                          That's the fencing feature required for the
                          engine to ensure that the<br>
                          host is actually offline. It won't resume any
                          other VMs to prevent<br>
                          potential VM corruption (eg. VM running on
                          multiple hosts).<br>
                          <br>
                          Andrew.<br>
                          <br>
                          On Jan 27, 2014 5:12 PM, "Jaison peter" &lt;<a
                            moz-do-not-send="true"
                            href="mailto:urotrip2@gmail.com"
                            target="_blank">urotrip2@gmail.com</a><br>
                          &lt;mailto:<a moz-do-not-send="true"
                            href="mailto:urotrip2@gmail.com"
                            target="_blank">urotrip2@gmail.com</a>&gt;&gt;
                          wrote:<br>
                          <br>
                          &nbsp; &nbsp; &nbsp;Hi all ,<br>
                          <br>
                          &nbsp; &nbsp; &nbsp;I was setting a two node ovirt cluster
                          with ovirt engine on<br>
                          &nbsp; &nbsp; &nbsp;seperate node . I completed the
                          configuration and tested VM &nbsp;live<br>
                          &nbsp; &nbsp; &nbsp;migrations with out any issues . Then for
                          checking cluster HA I<br>
                          &nbsp; &nbsp; &nbsp;powered down one host and expected vms
                          running on that host to be<br>
                          &nbsp; &nbsp; &nbsp;migrated to the other one . But nothing
                          happened , Engine detected<br>
                          &nbsp; &nbsp; &nbsp;host as un-rechable and marked it as
                          non-operational and vm ran on<br>
                          &nbsp; &nbsp; &nbsp;that host went to 'unknown state' . Is
                          that not possible to setup<br>
                          &nbsp; &nbsp; &nbsp;a fully HA ovirt cluster with two nodes ?
                          or else is that my<br>
                          &nbsp; &nbsp; &nbsp;configuration problem ? please advice .<br>
                          <br>
                          &nbsp; &nbsp; &nbsp;Thanks &amp; Regards<br>
                          <br>
                          &nbsp; &nbsp; &nbsp;Alex<br>
                          <br>
                          &nbsp; &nbsp; &nbsp;_______________________________________________<br>
                          &nbsp; &nbsp; &nbsp;Users mailing list<br>
                          &nbsp; &nbsp; &nbsp;<a moz-do-not-send="true"
                            href="mailto:Users@ovirt.org"
                            target="_blank">Users@ovirt.org</a>
                          &lt;mailto:<a moz-do-not-send="true"
                            href="mailto:Users@ovirt.org"
                            target="_blank">Users@ovirt.org</a>&gt;<br>
                          &nbsp; &nbsp; &nbsp;<a moz-do-not-send="true"
                            href="http://lists.ovirt.org/mailman/listinfo/users"
                            target="_blank">http://lists.ovirt.org/mailman/listinfo/users</a><br>
                          <br>
                          <br>
                          <br>
                          _______________________________________________<br>
                          Users mailing list<br>
                          <a moz-do-not-send="true"
                            href="mailto:Users@ovirt.org"
                            target="_blank">Users@ovirt.org</a><br>
                          <a moz-do-not-send="true"
                            href="http://lists.ovirt.org/mailman/listinfo/users"
                            target="_blank">http://lists.ovirt.org/mailman/listinfo/users</a><br>
                        </blockquote>
                        <br>
                        -- <br>
                        Dafna Ron<br>
                        _______________________________________________<br>
                        Users mailing list<br>
                        <a moz-do-not-send="true"
                          href="mailto:Users@ovirt.org" target="_blank">Users@ovirt.org</a><br>
                        <a moz-do-not-send="true"
                          href="http://lists.ovirt.org/mailman/listinfo/users"
                          target="_blank">http://lists.ovirt.org/mailman/listinfo/users</a><br>
                      </blockquote>
                      <br>
                      <br>
                    </blockquote>
                    <br>
                    <br>
                  </blockquote>
                  <br>
                </div>
              </div>
            </blockquote>
          </div>
          <br>
        </div>
      </div>
    </blockquote>
    <br>
  </body>
</html>