<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    On 5/18/2018 5:06 AM, Sylvain Bauza wrote:<br>
    <blockquote type="cite"
cite="mid:CALOCmumiuie+2L4FKYXJZFZ8mrrYrUaJtNmGs1wqD_byubsawQ@mail.gmail.com">
      <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
      <div dir="auto">
        <div><br>
          <br>
          <div class="gmail_quote">
            <div dir="ltr">Le ven. 18 mai 2018 à 13:59, Nadathur, Sundar
              <<a href="mailto:sundar.nadathur@intel.com"
                moz-do-not-send="true">sundar.nadathur@intel.com</a>>
              a écrit :<br>
            </div>
            <blockquote class="gmail_quote" style="margin:0 0 0
              .8ex;border-left:1px #ccc solid;padding-left:1ex">
              <div text="#000000" bgcolor="#FFFFFF">
                <p>Hi Matt,<br>
                </p>
                <div class="m_6016069294377656187moz-cite-prefix">On
                  5/17/2018 3:18 PM, Matt Riedemann wrote:<br>
                </div>
                <blockquote type="cite">On 5/17/2018 3:36 PM, Nadathur,
                  Sundar wrote: <br>
                  <blockquote type="cite">This applies only to the
                    resources that Nova handles, IIUC, which does not
                    handle accelerators. The generic method that Alex
                    talks about is obviously preferable but, if that is
                    not available in Rocky, is the filter an option? <br>
                  </blockquote>
                  <br>
                  If nova isn't creating accelerator resources managed
                  by cyborg, I have no idea why nova would be doing
                  quota checks on those types of resources. And no, I
                  don't think adding a scheduler filter to nova for
                  checking accelerator quota is something we'd add
                  either. I'm not sure that would even make sense - the
                  quota for the resource is per tenant, not per host is
                  it? The scheduler filters work on a per-host basis. <br>
                </blockquote>
                Can we not extend BaseFilter.filter_all() to get all the
                hosts in a filter? <br>
                          <a
                  class="m_6016069294377656187moz-txt-link-freetext"
                  href="https://github.com/openstack/nova/blob/master/nova/filters.py#L36"
                  target="_blank" rel="noreferrer"
                  moz-do-not-send="true">https://github.com/openstack/nova/blob/master/nova/filters.py#L36</a>
                <br>
                <br>
                I should have made it clearer that this putative filter
                will be out-of-tree, and needed only till better
                solutions become available.</div>
            </blockquote>
          </div>
        </div>
        <div dir="auto"><br>
        </div>
        <div dir="auto">No, there are two clear parameters for a filter,
          and changing that would mean a new paradigm for
          FilterScheduler.</div>
        <div dir="auto">If you need to have a check for all the hosts,
          maybe it should be either a pre-filter for Placement or a
          post-filter but we don't accept out of tree yet.</div>
      </div>
    </blockquote>
    <br>
    Thanks, Sylvain. So, the filter approach got filtered out.<br>
    <br>
    Matt had mentioned that Cinder volume quotas are not checked by Nova
    either, citing:<br>
         <a class="moz-txt-link-freetext" href="https://bugs.launchpad.net/nova/+bug/1742102">https://bugs.launchpad.net/nova/+bug/1742102</a> <br>
    That includes this comment:<br>
        <a class="moz-txt-link-freetext" href="https://bugs.launchpad.net/nova/+bug/1742102/comments/4">https://bugs.launchpad.net/nova/+bug/1742102/comments/4</a> <br>
    I'll check how Cinder does it today.<br>
    <br>
    Thanks to all for your valuable input.<br>
    <br>
    Regards,<br>
    Sundar<br>
  </body>
</html>