On Fri, Jan 9, 2009 at 16:25, Arndt Roth <[email protected]> wrote:
>
> You're right. Thanks.
> I had even chosen the correct name "*_meta_attr_*" and still wrote it in 
> instance_attributes.
>
> Fixed that and added globally_unique, but the crm_verify warning remained the 
> same.
> In the end a complete heartbeat restart of all nodes helped. So I guess it 
> was some stale information in the LRM?
>
> Here is the complete and now correct clone-with-group-of-OCF-and-LSB-resource 
> config:
>
>   <clone id="clone_ldirectord">
>     <meta_attributes id="clone_ldirectord-meta-options">
>       <attributes>
>         <nvpair id="clone_ldirector_conf_meta_attr_clone_max" 
> name="clone_max" value="3"/>
>         <nvpair id="clone_ldirector_conf_meta_attr_clone_node_max" 
> name="clone_node_max" value="1"/>

these two are wrong '_' -> '-'

and just changing the values wont be enough, you need to restart the
cluster in order to clear out the operation record

>         <nvpair id="clone_ldirector_conf_meta_attr_globally-unique" 
> name="globally-unique" value="false"/>
>         <nvpair id="clone_ldirectord-meta-options-target-role" 
> name="target-role" value="Started"/>
>         <nvpair id="clone_ldirectord-meta-options-is-managed" 
> name="is-managed" value="true"/>
>         <nvpair id="clone_ldirectord-meta-options-notify" name="notify" 
> value="true"/>
>       </attributes>
>     </meta_attributes>
>     <group id="group_clone_ldirectord_lvs-monitor">
>       <primitive id="resource_ldirectord" class="ocf" type="ldirectord" 
> provider="heartbeat">
>         <operations>
>           <op id="resource_ldirectord_operation_op" name="monitor" 
> description="ldirectord-monitor" interval="10" timeout="3" start_delay="10s" 
> disabled="false" role="Started" prereq="nothing" on_fail="restart"/>
>         </operations>
>         <instance_attributes id="ldirectord_inst_attributes">
>           <attributes>
>             <nvpair id="ldirector_attr_configfile" name="configfile" 
> value="/etc/ha.d/ldirectord.cf"/>
>             <nvpair id="ldirector_attr_binary" name="ldirectord" 
> value="/usr/sbin/ldirectord"/>
>           </attributes>
>         </instance_attributes>
>       </primitive>
>       <primitive class="lsb" id="lvs-monitor" type="lvs-monitor" 
> restart_type="restart">
>         <operations>
>           <op id="lvs-monitor_op_start" name="start" timeout="2s"/>
>           <op id="lvs-monitor_op_stop" name="stop" timeout="2s"/>
>           <op id="lvs-monitor_op_status" name="monitor" interval="5s" 
> timeout="2s"/>
>         </operations>
>       </primitive>
>     </group>
>   </clone>
>
>
> -----Ursprüngliche Nachricht-----
> Von: [email protected] 
> [mailto:[email protected]] Im Auftrag von Dejan Muhamedagic
> Gesendet: Freitag, 9. Januar 2009 15:30
> An: General Linux-HA mailing list
> Betreff: Re: [Linux-HA] crm_verify bug?
>
> On Fri, Jan 09, 2009 at 02:26:37PM +0100, Arndt Roth wrote:
>> Well, the clone runs on all nodes (see crm_mon output). That's fine and I 
>> guess it isn't globally unique then. It is just not verified correctly by 
>> crm_verify (I suppose).
>>
>> Here's the clone-config:
>>
>>
>>        <clone id="clone_ldirectord">
>>
>>          <instance_attributes id="clone_ldirectord_inst_attr">
>>            <attributes>
>>              <nvpair id="clone_ldirector_conf_meta_attr_clone_max" 
>> name="clone_max" value="3"/>
>>              <nvpair id="clone_ldirector_conf_meta_attr_clone_node_max" 
>> name="clone_node_max" value="1"/>
>
> Hence, you have to also set globally_unique to false here. BTW,
> shouldn't this be meta_attributes?
>
> Thanks,
>
> Dejan
>
>>            </attributes>
>>          </instance_attributes>
>>
>>          <group id="group_clone_ldirectord_lvs-monitor">
>>
>>            <primitive id="resource_ldirectord" class="ocf" type="ldirectord" 
>> provider="heartbeat">
>>              <operations>
>>                <op id="resource_ldirectord_operation_op" name="monitor" 
>> description="ldirectord-monitor" interval="10" timeout="3" start_delay="10s" 
>> disabled="false" role="Started" prereq="nothing" on_
>> fail="restart"/>
>>              </operations>
>>              <instance_attributes id="ldirectord_inst_attributes">
>>                <attributes>
>>                  <nvpair id="ldirector_attr_configfile" name="configfile" 
>> value="/etc/ha.d/ldirectord.cf"/>
>>                  <nvpair id="ldirector_attr_binary" name="ldirectord" 
>> value="/usr/sbin/ldirectord"/>
>>                </attributes>
>>              </instance_attributes>
>>            </primitive>
>>            <meta_attributes id="clone_ldirectord-meta-options">
>>              <attributes>
>>                <nvpair id="clone_ldirectord-meta-options-target-role" 
>> name="target-role" value="Started"/>
>>                <nvpair id="clone_ldirectord-meta-options-is-managed" 
>> name="is-managed" value="true"/>
>>                <nvpair id="clone_ldirectord-meta-options-notify" 
>> name="notify" value="true"/>
>>              </attributes>
>>            </meta_attributes>
>>
>>            <primitive class="lsb" id="lvs-monitor" type="lvs-monitor" 
>> restart_type="restart">
>>              <operations>
>>                <op id="lvs-monitor_op_start" name="start" timeout="2s"/>
>>                <op id="lvs-monitor_op_stop" name="stop" timeout="2s"/>
>>                <op id="lvs-monitor_op_status" name="monitor" interval="5s" 
>> timeout="2s"/>
>>              </operations>
>>            </primitive>
>>
>>          </group>
>>
>>        </clone>
>>
>> -----Urspr?ngliche Nachricht-----
>> Von: [email protected] 
>> [mailto:[email protected]] Im Auftrag von Dejan Muhamedagic
>> Gesendet: Freitag, 9. Januar 2009 13:04
>> An: General Linux-HA mailing list
>> Betreff: Re: [Linux-HA] crm_verify bug?
>>
>> Hi,
>>
>> On Fri, Jan 09, 2009 at 12:39:34PM +0100, Arndt Roth wrote:
>> > Hi *,
>> >
>> >
>> >
>> > I have a group with one OCF and one LSB resource defined into a clone
>> > (OCF-Clone + LSB-primitive in a group didn't work).
>> >
>> >
>> >
>> > Clone Set: clone_ldirectord
>> >
>> >     Resource Group: group_clone_ldirectord_lvs-monitor:0
>> >
>> >         resource_ldirectord:0   (ocf::heartbeat:ldirectord):    Started
>> > server2
>> >
>> >         lvs-monitor:0   (lsb:lvs-monitor):      Started server2
>> >
>> >     Resource Group: group_clone_ldirectord_lvs-monitor:1
>> >
>> >         resource_ldirectord:1   (ocf::heartbeat:ldirectord):    Started
>> > server1
>> >
>> >         lvs-monitor:1   (lsb:lvs-monitor):      Started server1
>> >
>> >     Resource Group: group_clone_ldirectord_lvs-monitor:2
>> >
>> >         resource_ldirectord:2   (ocf::heartbeat:ldirectord):    Started
>> > server3
>> >
>> >         lvs-monitor:2   (lsb:lvs-monitor):      Started server3
>> >
>> >
>> >
>> > Now I get this confusing error-message on all 3 nodes when verifying
>> > with crm_verify:
>> >
>> >
>> >
>> > [r...@server2:~]$ crm_verify -LV
>> >
>> > crm_verify[20534]: 2009/01/09_12:28:27 WARN: unpack_rsc_op:
>> > resource_ldirectord:0_monitor_0 found active resource_ldirectord:0 on
>> > server2
>> >
>> > crm_verify[20534]: 2009/01/09_12:28:27 WARN: unpack_rsc_op:
>> > resource_ldirectord:1_monitor_0 found active resource_ldirectord:1 on
>> > server2
>> >
>> > crm_verify[20534]: 2009/01/09_12:28:27 WARN: unpack_rsc_op:
>> > resource_ldirectord:2_monitor_0 found active resource_ldirectord:2 on
>> > server2
>> >
>> > crm_verify[20534]: 2009/01/09_12:28:27 WARN: unpack_rsc_op:
>> > lvs-monitor:2_monitor_0 found active lvs-monitor:2 on server2
>> >
>> > crm_verify[20534]: 2009/01/09_12:28:27 WARN: unpack_rsc_op:
>> > lvs-monitor:0_monitor_0 found active lvs-monitor:0 on server2
>> >
>> > crm_verify[20534]: 2009/01/09_12:28:27 WARN: unpack_rsc_op:
>> > lvs-monitor:1_monitor_0 found active lvs-monitor:1 on server2
>> >
>> >
>> >
>> > Why does crm_verify display on all nodes that all resources are active
>> > on server2?
>> >
>> > Apart from the error-message everything looks ok, but how can I confirm
>> > that it is only a crm_verify problem or bug?
>>
>> Perhaps you can show the configuration too. Maybe this is due to
>> globally-unique set to true (the default) for the clone?
>>
>> Thanks,
>>
>> Dejan
>>
>> >
>> >
>> > Thanks for advice,
>> >
>> >
>> >
>> > Arndt
>> >
>> >
>> >
>> >
>> >
>> >
>> > _______________________________________________
>> > Linux-HA mailing list
>> > [email protected]
>> > http://lists.linux-ha.org/mailman/listinfo/linux-ha
>> > See also: http://linux-ha.org/ReportingProblems
>> _______________________________________________
>> Linux-HA mailing list
>> [email protected]
>> http://lists.linux-ha.org/mailman/listinfo/linux-ha
>> See also: http://linux-ha.org/ReportingProblems
>>
>>
>>
>> _______________________________________________
>> Linux-HA mailing list
>> [email protected]
>> http://lists.linux-ha.org/mailman/listinfo/linux-ha
>> See also: http://linux-ha.org/ReportingProblems
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
>
>
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to