[Pacemaker] target-role changing from Master to Started

Rick Kint rwkint at yahoo.com
Thu Nov 18 01:08:45 UTC 2010


Hello --

We've experienced a weird problem 3 times in the last month on different clusters.

Environment:
- Pacemaker 1.0.9.1-1
- Corosync 1.2.5-1
- RHEL5.4
- 2 nodes

We run a simple configuration with a master/slave resource, appended below.

Three times we have found different clusters in a state where both instances are slaves and Pacemaker seems perfectly happy with this.  Looking at the configuration (cib.xml) reveals that the target-role attribute for the master/slave resource has changed from Master to Started.  Changing the value of this attribute back to Master fixes the problem immediately.

Any ideas what might be causing this?  Nobody is doing it manually.

I'll hold off on logs until asked, this stuff can get voluminous.


Thanks, Rick Kint


<configuration>
  <crm_config>
    <cluster_property_set id="cib-bootstrap-options">
      <nvpair id="cib-bootstrap-options-stonith-enabled" name="stonith-enabled" value="false"/>
      <nvpair id="cib-bootstrap-options-no-quorum-policy" name="no-quorum-policy" value="ignore"/>
      <nvpair id="cib-bootstrap-options-dc-version" name="dc-version" value="1.0.9-89bd754939df5150de7cd76835f98fe90851b677"/>
      <nvpair id="cib-bootstrap-options-cluster-infrastructure" name="cluster-infrastructure" value="openais"/>
      <nvpair id="cib-bootstrap-options-expected-quorum-votes" name="expected-quorum-votes" value="2"/>
      <nvpair id="cib-bootstrap-options-last-lrm-refresh" name="last-lrm-refresh" value="1289857566"/>
    </cluster_property_set>
  </crm_config>
  <nodes>
    <node id="fnord004" type="normal" uname="fnord004"/>
    <node id="fnord005" type="normal" uname="fnord005"/>
  </nodes>
  <resources>
    <master id="Bar">
      <meta_attributes id="Bar-meta_attributes">
        <nvpair id="Bar-meta_attributes-clone-max" name="clone-max" value="2"/>
        <nvpair id="Bar-meta_attributes-clone-node-max" name="clone-node-max" value="1"/>
        <nvpair id="Bar-meta_attributes-master-max" name="master-max" value="1"/>
        <nvpair id="Bar-meta_attributes-notify" name="notify" value="false"/>
        <nvpair id="Bar-meta_attributes-target-role" name="target-role" value="Master"/>
        <nvpair id="Bar-meta_attributes-is-managed" name="is-managed" value="true"/>
      </meta_attributes>
      <primitive class="ocf" id="Foo" provider="fnord" type="foo">
        <operations>
          <op id="Foo-start-0s" interval="0s" name="start" timeout="300s"/>
          <op id="Foo-monitor-2s" interval="2s" name="monitor" role="Master" timeout="20s"/>
          <op id="Foo-monitor-3s" interval="3s" name="monitor" role="Slave" timeout="20s"/>
        </operations>
      </primitive>
    </master>
  </resources>
  <constraints/>
  <op_defaults/>
  <rsc_defaults>
    <meta_attributes id="rsc-options">
      <nvpair id="rsc-options-resource-stickiness" name="resource-stickiness" value="100"/>
    </meta_attributes>
  </rsc_defaults>
</configuration>






      
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://oss.clusterlabs.org/pipermail/pacemaker/attachments/20101117/89aeaf1b/attachment-0001.htm>


More information about the Pacemaker mailing list