[ClusterLabs] Reason for changing a meta_attribute in the CIB

[Ken Gaillot]

On 07/01/2015 04:31 AM, Detlef Gossrau wrote:
> Hi all,
> 
> how can I find the reason for changing a meta_attribute of a lsb
> resource in the CIB ? All what I see in /var/log/messages is
> 
> Jun 25 11:51:10 muc-ant-02 cib[1210]:     info: cib:diff: +
> <configuration >
> Jun 25 11:51:10 muc-ant-02 cib[1210]:     info: cib:diff: +    
> <resources >
> Jun 25 11:51:10 muc-ant-02 cib[1210]:     info: cib:diff: + <primitive
> class="lsb" id="resRecorder" type="andphone" >
> Jun 25 11:51:10 muc-ant-02 cib[1210]:     info: cib:diff: +
> <meta_attributes id="resRecorder-meta_attributes" >
> Jun 25 11:51:10 muc-ant-02 cib[1210]:     info: cib:diff: + <nvpair
> id="resRecorder-meta_attributes-target-role" name="target-role"
> value="Stopped" __crm_diff_marker__="added:top" />
> Jun 25 11:51:10 muc-ant-02 cib[1210]:     info: cib:diff: +
> </meta_attributes>
> Jun 25 11:51:10 muc-ant-02 cib[1210]:     info: cib:diff: +
> <meta_attributes id="resRecorder-meta_attributes-0" >
> Jun 25 11:51:10 muc-ant-02 cib[1210]:     info: cib:diff: + <nvpair
> id="resRecorder-meta_attributes-0-target-role" name="target-role"
> value="Stopped" __crm_diff_marker__="added:top" />
> Jun 25 11:51:10 muc-ant-02 cib[1210]:     info: cib:diff: +
> </meta_attributes>
> Jun 25 11:51:10 muc-ant-02 crmd[1215]:     info: abort_transition_graph:
> te_update_diff:126 - Triggered transition abort (complete=1, tag=diff,
> id=(null), magic=NA, cib=0.118.1) : Non-status change
> Jun 25 11:51:10 muc-ant-02 crmd[1215]:   notice: do_state_transition:
> State transition S_IDLE -> S_POLICY_ENGINE [ input=I_PE_CALC
> cause=C_FSA_INTERNAL origin=abort_transition_graph ]
> Jun 25 11:51:10 muc-ant-02 cib[1210]:     info: cib:diff: + </primitive>
> Jun 25 11:51:10 muc-ant-02 cib[1210]:     info: cib:diff: +    
> </resources>
> Jun 25 11:51:10 muc-ant-02 cib[1210]:     info: cib:diff: +
> </configuration>
> Jun 25 11:51:10 muc-ant-02 cib[1210]:     info: cib:diff: + </cib>
> 
> After that the ressource is stopped.

A change can be initiated by any node; the other nodes will be notified
of the change, which you're seeing above. You'll have to look at the
logs around that time on each of the other nodes to figure out which one
initiated it and why. Whichever one was DC at the time is the most
likely initiator.

> Is there a possibility to set a debug level to get more information ? I
> can't find any hint or dependency why the meta_attribute is changed.

Yes, but most likely you just need to check the logs on the DC.

Typically, pacemaker generates two log streams, one to syslog and one to
/var/log/pacemaker.log (specifics can vary by Linux distribution and
your configuration). The pacemaker.log will be more detailed, but you
typically have to code-dive to make sense of it.

In a testing environment, you can turn up the verbosity, but it can
quickly become overwhelming. Typically these settings will be in
/etc/sysconfig/pacemaker or /etc/default/pacemaker. There are a number
of variables that can be set there, for example PCMK_debug=yes, but be
prepared to give your disk some good exercise.

> Test environment on both nodes:
> 
> OS         : Linux 2.6.32-504.8.1.el6.x86_64
> Cluster SW : pacemaker-cluster-libs-1.1.7-6.el6.x86_64
>              pacemaker-libs-1.1.7-6.el6.x86_64
>              pacemaker-cli-1.1.7-6.el6.x86_64
>              pacemaker-1.1.7-6.el6.x86_64
>              corosynclib-1.4.1-17.el6.x86_64
>              corosync-1.4.1-17.el6.x86_64

For a simple cluster, those versions should be fine, but be aware that
they're considered old, and fewer people will be able to give help here
or on freenode's #clusterlabs channel. Most likely someone will tell you
to upgrade first. :)

>From the versions, I'm guessing you're using Debian; there's an effort
underway to get recent versions in sid and from there into backports.
But most people use RHEL/CentOS, SLES or Ubuntu because historically
those have had newer versions and better support.

> Thnx for any help !
> 
> Best regards,
> Detlef