Dev: Parametrize the log dir

[aleksei-burlakov]

By default all logs are stored in /var/lib/heartbeat/trace_ra
This patch adds a new resource attribute trace_dir to the cib.file
The trace_dir should come alongside the trace_ra,
whereas trace_ra signals that the resource is traced,
the trace_dir defines the path.

[jpbuecken]

Nice, as a user, this is what we needed (as you know, a cluster is used to mitigate downtimes, and if we need to enable tracing in production a full /var is not very good to avoid downtimes :-) ).

Question: Will trace_dir be available as cluster wide default option so we can add it to our default setup and only switch trace_ra per resource if needed?

e.g

rsc_defaults rsc-options: \
        resource-stickiness=100 \
        is-managed=true \
        trace_dir="/extra/trace/fs"

EDIT:
or as xml/cib

    <rsc_defaults>
      <meta_attributes id="rsc-options">
        <nvpair name="resource-stickiness" value="100" id="rsc-options-resource-stickiness"/>
        <nvpair name="is-managed" value="true" id="rsc-options-is-managed"/>
        <nvpair name="trace_dir" value="/extra/trace/fs" id="rsc-options-trace_dir"/>
      </meta_attributes>
    </rsc_defaults>

[aleksei-burlakov]

Oops, I forgot to put the reference to the draft PR #935

@jpbuecken , exactly, the cib file will be changed like (assumung there is a Dummy resource dummy1)

     <resources>                                                                                                                                                                             
       <primitive id="dummy1" class="ocf" provider="pacemaker" type="Dummy">                                                                                                                 
         <operations>                                                                                                                                                                                                    
          <op name="monitor" interval="15" id="dummy1-monitor-15"/>
           <op name="start" interval="0" id="dummy1-start-0">
             <instance_attributes id="dummy1-start-0-instance_attributes">
                <nvpair name="trace_ra" value="1" id="dummy1-start-0-instance_attributes-trace_ra"/>
+               <nvpair name="trace_dir" value="/foo/bar" id="dummy1-start-0-instance_attributes-trace_dir"/>
              </instance_attributes>                                                           
           </op>                                                                              
         </operations>                                                                        
       </primitive>                                                                           
     </resources>

[gao-yan]

Question: Will trace_dir be available as cluster wide default option so we can add it to our default setup and only switch trace_ra per resource if needed?

It's an op parameter, so it's likely to work with op_defaults like:

op_defaults trace_dir="/extra/trace/fs"

[aleksei-burlakov]

Hi @liangxin1300 , can you please approve my PR?

[liangxin1300]

Hi @liangxin1300 , can you please approve my PR?

Sorry for late, I will take a look today

Thanks!

[liangxin1300]

Other part of codes looks good to me.
Nice work! @aleksei-burlakov
After changing the doc part, I will merge and submit the code into 15SP4 in next week