[Pacemaker] Asymmetrical cluster : should it probe resources that won't be installed on a node ?

Thomas Guthmann tguthmann at iseek.com.au
Tue Dec 8 00:14:44 UTC 2009


Hi,

I was wondering why pacemaker is throwing errors about program not 
installed when it is normal that is not. For instance, on a loadbalancer 
node (ed-lb2) we don't want to host named (DNS daemon). The cluster is 
working so far even if pacemaker seems not happy because it still tries 
to see if it can monitor named on nodes that don't and won't have named 
(reminder: we have symmetric-cluster=false :)

I don't know if our cluster is missing some configuration parameters to 
avoid that or if it's a pacemaker limitation or bug. BTW, as a side 
effect, we also have a lot of "pe-warn" files filling our disk space :)

Any ideas ?

Cheers,
Thomas

---

OS : Linux Centos 5.4
Packages :
   pacemaker-1.0.5-4.1
   pacemaker-mgmt-1.99.2-6.1
   pacemaker-libs-1.0.5-4.1
   libopenais2-0.80.5-15.1
   openais-0.80.5-15.1

--------------------------------------------------------------------

[ed-lb2:~] # crm_verify -L
crm_verify[2251]: 2009/12/08_09:37:47 ERROR: unpack_rsc_op: Hard error - 
ed-named:0_monitor_0 failed with rc=5: Preventing ed-dns from 
re-starting on ed-lb2.internal

Error 5 is "program is not installed". The logs in messages are even 
more explicit :

--------------------------------------------------------------------

Dec  8 09:50:07 ed-lb2 crmd: [7034]: info: do_lrm_rsc_op: Performing 
key=25:16453:7:349cea8f-8086-4f75-85f2-9d56d1b87989 
op=ed-named:0_monitor_0 )
Dec  8 09:50:07 ed-lb2 lrmd: [7031]: info: rsc:ed-named:0:8239: monitor
Dec  8 09:50:07 ed-lb2 named[11022]: ERROR: Can't find rndc binary at 
'/usr/sbin/rndc', must specify it's location in resource attribute 
'rndcbin'.
Dec  8 09:50:07 ed-lb2 lrmd: [7031]: info: RA output: 
(ed-named:0:monitor:stderr) 2009/12/08_09:50:07 ERROR: Can't find rndc 
binary at '/usr/sbin/rndc', must specify it's location in resource 
attribute 'rndcbin'.
Dec  8 09:50:07 ed-lb2 crmd: [7034]: info: process_lrm_event: LRM 
operation ed-named:0_monitor_0 (call=8239, rc=5, cib-update=8254, 
confirmed=true) complete not installed

--------------------------------------------------------------------

       <clone id="ed-dns">
         <meta_attributes id="ed-dns-meta_attributes">
           <nvpair id="ed-dns-meta_attributes-clone-max" 
name="clone-max" value="2"/>
           <nvpair id="ed-dns-meta_attributes-target-role" 
name="target-role" value="started"/>
         </meta_attributes>
         <primitive class="ocf" id="ed-named" provider="iseek" type="named">
           <operations id="ed-named-operations">
             <op id="ed-named-monitor-5" interval="5" name="monitor" 
start-delay="0" timeout="20">
               <instance_attributes 
id="ed-named-monitor-5-instance_attributes">
                 <nvpair 
id="ed-named-monitor-5-instance_attributes-OCF_CHECK_LEVEL" 
name="OCF_CHECK_LEVEL" value="10"/>
               </instance_attributes>
             </op>
           </operations>
           <meta_attributes id="ed-named-meta_attributes"/>
         </primitive>
       </clone>

--------------------------------------------------------------------






More information about the Pacemaker mailing list