<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  </head>
  <body>
    <div class="moz-cite-prefix">On 12/14/20 11:48 AM, Gabriele Bulfon
      wrote:<br>
    </div>
    <blockquote type="cite" cite="mid:1065144646.7212.1607942889206@www">
      <meta http-equiv="content-type" content="text/html; charset=UTF-8">
      <header></header>
      <div style="font-family: tahoma,arial,helvetica,sans-serif;
        font-size: 14px;">Thanks!<br>
        <br>
        I tried first option, by adding pcmk_delay_base to the two
        stonith primitives.</div>
      <div style="font-family: tahoma,arial,helvetica,sans-serif;
        font-size: 14px;">First has 1 second, second has 5 seconds.</div>
      <div style="font-family: tahoma,arial,helvetica,sans-serif;
        font-size: 14px;">It didn't work :( they still killed each other
        :(</div>
      <div style="font-family: tahoma,arial,helvetica,sans-serif;
        font-size: 14px;">Anything wrong with the way I did it?</div>
    </blockquote>
    <tt>Maybe 4s difference is a bit close.</tt><tt><br>
    </tt><tt>Did you try anything significantly larger?</tt><tt><br>
    </tt><tt><br>
    </tt><tt>Klaus</tt><br>
    <blockquote type="cite" cite="mid:1065144646.7212.1607942889206@www">
      <div style="font-family: tahoma,arial,helvetica,sans-serif;
        font-size: 14px;"> </div>
      <div style="font-family: tahoma,arial,helvetica,sans-serif;
        font-size: 14px;">Here's the config:</div>
      <div style="font-family: tahoma,arial,helvetica,sans-serif;
        font-size: 14px;"> </div>
      <div style="font-family: tahoma,arial,helvetica,sans-serif;
        font-size: 14px;"><span style="font-family: 'courier new',
          courier;">node 1: xstha1 \</span><br>
        <span style="font-family: 'courier new', courier;">       
          attributes standby=off maintenance=off</span><br>
        <span style="font-family: 'courier new', courier;">node 2:
          xstha2 \</span><br>
        <span style="font-family: 'courier new', courier;">       
          attributes standby=off maintenance=off</span><br>
        <span style="font-family: 'courier new', courier;">primitive
          xstha1-stonith stonith:external/ipmi \</span><br>
        <span style="font-family: 'courier new', courier;">       
          params hostname=xstha1 ipaddr=192.168.221.18 userid=ADMIN
          passwd="***" interface=lanplus pcmk_delay_base=1 \</span><br>
        <span style="font-family: 'courier new', courier;">        op
          monitor interval=25 timeout=25 start-delay=25 \</span><br>
        <span style="font-family: 'courier new', courier;">        meta
          target-role=Started</span><br>
        <span style="font-family: 'courier new', courier;">primitive
          xstha1_san0_IP IPaddr \</span><br>
        <span style="font-family: 'courier new', courier;">       
          params ip=10.10.10.1 cidr_netmask=255.255.255.0 nic=san0</span><br>
        <span style="font-family: 'courier new', courier;">primitive
          xstha2-stonith stonith:external/ipmi \</span><br>
        <span style="font-family: 'courier new', courier;">       
          params hostname=xstha2 ipaddr=192.168.221.19 userid=ADMIN
          passwd="***" interface=lanplus pcmk_delay_base=5 \</span><br>
        <span style="font-family: 'courier new', courier;">        op
          monitor interval=25 timeout=25 start-delay=25 \</span><br>
        <span style="font-family: 'courier new', courier;">        meta
          target-role=Started</span><br>
        <span style="font-family: 'courier new', courier;">primitive
          xstha2_san0_IP IPaddr \</span><br>
        <span style="font-family: 'courier new', courier;">       
          params ip=10.10.10.2 cidr_netmask=255.255.255.0 nic=san0</span><br>
        <span style="font-family: 'courier new', courier;">primitive
          zpool_data ZFS \</span><br>
        <span style="font-family: 'courier new', courier;">       
          params pool=test \</span><br>
        <span style="font-family: 'courier new', courier;">        op
          start timeout=90 interval=0 \</span><br>
        <span style="font-family: 'courier new', courier;">        op
          stop timeout=90 interval=0 \</span><br>
        <span style="font-family: 'courier new', courier;">        meta
          target-role=Started</span><br>
        <span style="font-family: 'courier new', courier;">location
          xstha1-stonith-pref xstha1-stonith -inf: xstha1</span><br>
        <span style="font-family: 'courier new', courier;">location
          xstha1_san0_IP_pref xstha1_san0_IP 100: xstha1</span><br>
        <span style="font-family: 'courier new', courier;">location
          xstha2-stonith-pref xstha2-stonith -inf: xstha2</span><br>
        <span style="font-family: 'courier new', courier;">location
          xstha2_san0_IP_pref xstha2_san0_IP 100: xstha2</span><br>
        <span style="font-family: 'courier new', courier;">order
          zpool_data_order inf: zpool_data ( xstha1_san0_IP )</span><br>
        <span style="font-family: 'courier new', courier;">location
          zpool_data_pref zpool_data 100: xstha1</span><br>
        <span style="font-family: 'courier new', courier;">colocation
          zpool_data_with_IPs inf: zpool_data xstha1_san0_IP</span><br>
        <span style="font-family: 'courier new', courier;">property
          cib-bootstrap-options: \</span><br>
        <span style="font-family: 'courier new', courier;">       
          have-watchdog=false \</span><br>
        <span style="font-family: 'courier new', courier;">       
          dc-version=1.1.15-e174ec8 \</span><br>
        <span style="font-family: 'courier new', courier;">       
          cluster-infrastructure=corosync \</span><br>
        <span style="font-family: 'courier new', courier;">       
          stonith-action=poweroff \</span><br>
        <span style="font-family: 'courier new', courier;">       
          no-quorum-policy=stop</span></div>
      <div style="font-family: tahoma,arial,helvetica,sans-serif;
        font-size: 14px;"> </div>
      <div id="wt-mailcard">
        <div> </div>
        <div><span style="font-size: 14px; font-family: Helvetica;"><strong>Sonicle
              S.r.l. </strong>: <a href="http://www.sonicle.com/"
              target="_new" moz-do-not-send="true">http://www.sonicle.com</a></span></div>
        <div><span style="font-size: 14px; font-family: Helvetica;"><strong>Music: </strong><a
              href="http://www.gabrielebulfon.com/" target="_new"
              moz-do-not-send="true">http://www.gabrielebulfon.com</a></span></div>
        <div><span style="font-size: 14px; font-family: Helvetica;"><strong>eXoplanets : </strong><a
href="https://gabrielebulfon.bandcamp.com/album/exoplanets"
              moz-do-not-send="true">https://gabrielebulfon.bandcamp.com/album/exoplanets</a></span></div>
        <div> </div>
      </div>
      <div style="font-family: tahoma,arial,helvetica,sans-serif;
        font-size: 14px;"><tt><br>
          <br>
          <br>
----------------------------------------------------------------------------------<br>
          <br>
          Da: Andrei Borzenkov <a class="moz-txt-link-rfc2396E" href="mailto:arvidjaar@gmail.com"><arvidjaar@gmail.com></a><br>
          A: <a class="moz-txt-link-abbreviated" href="mailto:users@clusterlabs.org">users@clusterlabs.org</a> <br>
          Data: 13 dicembre 2020 7.50.57 CET<br>
          Oggetto: Re: [ClusterLabs] Antw: [EXT] Recoveing from node
          failure<br>
          <br>
        </tt></div>
      <blockquote style="border-left: #000080 2px solid; margin-left:
        5px; padding-left: 5px;"><tt>12.12.2020 20:30, Gabriele Bulfon
          пишет:<br>
          > Thanks, I will experiment this.<br>
          >  <br>
          > Now, I have a last issue about stonith.<br>
          > I tried to reproduce a stonith situation, by disabling
          the network interface used for HA on node 1.<br>
          > Stonith is configured with ipmi poweroff.<br>
          > What happens, is that once the interface is down, both
          nodes tries to stonith the other node, causing both to
          poweroff...<br>
          <br>
          Yes, this is expected. The options are basically<br>
          <br>
          1. Have separate stonith resource for each node and configure
          static<br>
          (pcmk_delay_base) or random dynamic (pcmk_delay_max) delays to
          avoid<br>
          both nodes starting stonith at the same time. This does not
          take<br>
          resources in account.<br>
          <br>
          2. Use fencing topology and create pseudo-stonith agent that
          does not<br>
          attempt to do anything but just delays for some time before
          continuing<br>
          with actual fencing agent. Delay can be based on anything
          including<br>
          resources running on node.<br>
          <br>
          3. If you are using pacemaker 2.0.3+, you could use new<br>
          priority-fencing-delay feature that implements resource-based
          priority<br>
          fencing:<br>
          <br>
          + controller/fencing/scheduler: add new feature
          'priority-fencing-delay'<br>
          Optionally derive the priority of a node from the<br>
          resource-priorities<br>
          of the resources it is running.<br>
          In a fencing-race the node with the highest priority has a
          certain<br>
          advantage over the others as fencing requests for that node
          are<br>
          executed with an additional delay.<br>
          controlled via cluster option priority-fencing-delay (default
          = 0)<br>
          <br>
          <br>
          See also
          <a class="moz-txt-link-freetext" href="https://www.mail-archive.com/users@clusterlabs.org/msg10328.html">https://www.mail-archive.com/users@clusterlabs.org/msg10328.html</a><br>
          <br>
          > I would like the node running all resources (zpool and
          nfs ip) to be the first trying to stonith the other node.<br>
          > Or is there anything else better?<br>
          >  <br>
          > Here is the current crm config show:<br>
          >  <br>
          <br>
          It is unreadable<br>
          <br>
          > node 1: xstha1 \ attributes standby=off
          maintenance=offnode 2: xstha2 \ attributes standby=off
          maintenance=offprimitive xstha1-stonith stonith:external/ipmi
          \ params hostname=xstha1 ipaddr=192.168.221.18 userid=ADMIN
          passwd="******" interface=lanplus \ op monitor interval=25
          timeout=25 start-delay=25 \ meta target-role=Startedprimitive
          xstha1_san0_IP IPaddr \ params ip=10.10.10.1
          cidr_netmask=255.255.255.0 nic=san0primitive xstha2-stonith
          stonith:external/ipmi \ params hostname=xstha2
          ipaddr=192.168.221.19 userid=ADMIN passwd="******"
          interface=lanplus \ op monitor interval=25 timeout=25
          start-delay=25 \ meta target-role=Startedprimitive
          xstha2_san0_IP IPaddr \ params ip=10.10.10.2
          cidr_netmask=255.255.255.0 nic=san0primitive zpool_data ZFS \
          params pool=test \ op start timeout=90 interval=0 \ op stop
          timeout=90 interval=0 \ meta target-role=Startedlocation
          xstha1-stonith-pref xstha1-stonith -inf: xstha1location
          xstha1_san0_IP_pref xstha1_san0_IP 100: xstha1location
          xstha2-stonith-pref xstha2-stonith -inf: xstha2location
          xstha2_san0_IP_pref xstha2_san0_IP 100: xstha2order
          zpool_data_order inf: zpool_data ( xstha1_san0_IP )location
          zpool_data_pref zpool_data 100: xstha1colocation
          zpool_data_with_IPs inf: zpool_data xstha1_san0_IPproperty
          cib-bootstrap-options: \ have-watchdog=false \
          dc-version=1.1.15-e174ec8 \ cluster-infrastructure=corosync \
          stonith-action=poweroff \ no-quorum-policy=stop<br>
          >  <br>
          > Thanks!<br>
          > Gabriele<br>
          >  <br>
          >  <br>
          > Sonicle S.r.l. : <a class="moz-txt-link-freetext" href="http://www.sonicle.com">http://www.sonicle.com</a><br>
          > Music: <a class="moz-txt-link-freetext" href="http://www.gabrielebulfon.com">http://www.gabrielebulfon.com</a><br>
          >
          eXoplanets : <a class="moz-txt-link-freetext" href="https://gabrielebulfon.bandcamp.com/album/exoplanets">https://gabrielebulfon.bandcamp.com/album/exoplanets</a><br>
          >  <br>
          > <br>
          > <br>
          > <br>
          > <br>
          >
----------------------------------------------------------------------------------<br>
          > <br>
          > Da: Andrei Borzenkov <a class="moz-txt-link-rfc2396E" href="mailto:arvidjaar@gmail.com"><arvidjaar@gmail.com></a><br>
          > A: <a class="moz-txt-link-abbreviated" href="mailto:users@clusterlabs.org">users@clusterlabs.org</a> <br>
          > Data: 11 dicembre 2020 18.30.29 CET<br>
          > Oggetto: Re: [ClusterLabs] Antw: [EXT] Recoveing from
          node failure<br>
          > <br>
          > <br>
          > 11.12.2020 18:37, Gabriele Bulfon пишет:<br>
          >> I found I can do this temporarily:<br>
          >>  <br>
          >> crm config property cib-bootstrap-options:
          no-quorum-policy=ignore<br>
          >>  <br>
          > <br>
          > All two node clusters I remember run with setting forever
          :)<br>
          > <br>
          >> then once node 2 is up again:<br>
          >>  <br>
          >> crm config property cib-bootstrap-options:
          no-quorum-policy=stop<br>
          >>  <br>
          >> so that I make sure nodes will not mount in another
          strange situation.<br>
          >>  <br>
          >> Is there any better way? <br>
          > <br>
          > "better" us subjective, but ...<br>
          > <br>
          >> (such as ignore until everything is back to normal
          then conisder top again)<br>
          >>  <br>
          > <br>
          > That is what stonith does. Because quorum is pretty much
          useless in two<br>
          > node cluster, as I already said all clusters I have seem
          used<br>
          > no-quorum-policy=ignore and stonith-enabled=true. It
          means when node<br>
          > boots and other node is not available stonith is
          attempted; if stonith<br>
          > succeeds pacemaker continues with starting resources; if
          stonith fails,<br>
          > node is stuck.<br>
          > <br>
          > _______________________________________________<br>
          > Manage your subscription:<br>
          > <a class="moz-txt-link-freetext" href="https://lists.clusterlabs.org/mailman/listinfo/users">https://lists.clusterlabs.org/mailman/listinfo/users</a><br>
          > <br>
          > ClusterLabs home: <a class="moz-txt-link-freetext" href="https://www.clusterlabs.org/">https://www.clusterlabs.org/</a><br>
          > <br>
          > <br>
          > <br>
          > <br>
          > _______________________________________________<br>
          > Manage your subscription:<br>
          > <a class="moz-txt-link-freetext" href="https://lists.clusterlabs.org/mailman/listinfo/users">https://lists.clusterlabs.org/mailman/listinfo/users</a><br>
          > <br>
          > ClusterLabs home: <a class="moz-txt-link-freetext" href="https://www.clusterlabs.org/">https://www.clusterlabs.org/</a><br>
          > <br>
          <br>
          _______________________________________________<br>
          Manage your subscription:<br>
          <a class="moz-txt-link-freetext" href="https://lists.clusterlabs.org/mailman/listinfo/users">https://lists.clusterlabs.org/mailman/listinfo/users</a><br>
          <br>
          ClusterLabs home: <a class="moz-txt-link-freetext" href="https://www.clusterlabs.org/">https://www.clusterlabs.org/</a><br>
          <br>
          <br>
        </tt></blockquote>
      <br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <pre class="moz-quote-pre" wrap="">_______________________________________________
Manage your subscription:
<a class="moz-txt-link-freetext" href="https://lists.clusterlabs.org/mailman/listinfo/users">https://lists.clusterlabs.org/mailman/listinfo/users</a>

ClusterLabs home: <a class="moz-txt-link-freetext" href="https://www.clusterlabs.org/">https://www.clusterlabs.org/</a>
</pre>
    </blockquote>
    <br>
  </body>
</html>