[Pacemaker] Cluster with fc18 and fc17 nodes.

[Francis SOUYRI]

Hello Andrew,

Thank you for your reply...

I set this corosync config:

totem {
version: 2
secauth: off
cluster_name: cluster
   interface {
         ringnumber: 0
         bindnetaddr: 192.168.1.0
         ttl: 1
   }
transport: udpu
}

nodelist {
   node {
         ring0_addr: noeud1.xxxx.fr
         nodeid: 1
        }
   node {
         ring0_addr: noeud2.xxxx.fr
         nodeid: 2
        }
}

quorum {
provider: corosync_votequorum
}

logging {
to_syslog: yes
debug: off
}

Start the cluster, all ok (?), stop/start the cluster now I have this:

on FC18:
     <nodes>
       <node id="-1062731267" uname="noeud1.xxxx.fr" type="normal"/>
       <node id="-33445696" uname="noeud2.xxxx.fr" type="normal"/>
       <node id="3232236029" uname="noeud1.xxxx.fr"/>
       <node id="4261521600" uname="noeud2.xxxx.fr"/>
       <node id="1" uname="noeud1.xxxx.fr"/>
       <node id="2" uname="noeud2.xxxx.fr"/>
     </nodes>

...
Aug 26 10:38:39 noeud1 cib[8505]:  warning: cib_process_replace: 
Replacement 0.5.4 from noeud2.xxxx.fr not applied to 0.9.1: current 
epoch is greater than the replacement
...

on FC17:
     <nodes>
       <node id="-1062731267" uname="noeud1.xxxx.fr" type="normal"/>
       <node id="-33445696" uname="noeud2.xxxx.fr" type="normal"/>
     </nodes>

...
Aug 26 10:32:55 noeud2 crmd[23373]:   notice: do_state_transition: State 
transition S_INTEGRATION -> S_FINALIZE_JOIN [ input=I_INTEGRATED 
cause=C_FSA_INTERNAL origin=check_join_state ]
Aug 26 10:32:55 noeud2 crmd[23373]:     info: do_dc_join_finalize: 
join-14301: Syncing the CIB from noeud1.apec.fr to the rest of the cluster
Aug 26 10:32:55 noeud2 cib[23368]:    error: cib_perform_op: Discarding 
update with feature set '3.0.7' greater than our own '3.0.6'
Aug 26 10:32:55 noeud2 cib[23368]:  warning: cib_diff_notify: Update 
(client: crmd, call:28613): -1.-1.-1 -> 0.9.1 (The action/feature is not 
supported)
Aug 26 10:32:55 noeud2 cib[23368]:    error: cib_process_request: 
Operation complete: op cib_replace for section 'all' 
(origin=noeud1.xxxx.fr/crmd/28613, version=0.5.4): The action/feature is 
not supported (rc=-29)
Aug 26 10:32:55 noeud2 cib[23368]:     info: cib_process_request: 
Operation complete: op cib_sync for section 'all' 
(origin=noeud1.xxxx.fr/noeud1.xxxx.fr/28613, version=0.5.4): ok (rc=0)
Aug 26 10:32:55 noeud2 crmd[23373]:    error: finalize_sync_callback: 
Sync from noeud1.xxxx.fr failed: The action/feature is not supported
Aug 26 10:32:55 noeud2 crmd[23373]:  warning: do_log: FSA: Input 
I_ELECTION_DC from finalize_sync_callback() received in state 
S_FINALIZE_JOIN
Aug 26 10:32:55 noeud2 crmd[23373]:   notice: do_state_transition: State 
transition S_FINALIZE_JOIN -> S_INTEGRATION [ input=I_ELECTION_DC 
cause=C_FSA_INTERNAL origin=finalize_sync_callback ]
Aug 26 10:32:55 noeud2 crmd[23373]:     info: do_dc_join_offer_all: 
join-14302: Waiting on 2 outstanding join acks
Aug 26 10:32:55 noeud2 crmd[23373]:     info: update_dc: Set DC to 
noeud2.xxxx.fr (3.0.6)
...


Best regards.

Francis

On 08/26/2013 01:42 AM, Andrew Beekhof wrote:
>
> On 23/08/2013, at 7:18 PM, Francis SOUYRI <francis.souyri at apec.fr> wrote:
>
>> Hello,
>>
>> For a long time I used heartbeat/drbd for 2 nodes clusters with Fedora, I used the internal crm of heartbeat not pacemaker.
>>
>> I planned to upgrade from the fc17 to the fc18, but on fc18 heartbeat is obsolete and I have to change to corosync/pacemaker.
>> For information the heartbeat fc17 package work fine on fc18 and the cluster with a node fc17 and the other fc18 (without the firewall activated by default !!!) work perfectly (The final configuration is to have the both node in fc18).
>>
>> But the corosync/pacemaker do not work with a fc17 node and a fc18 node.
>>
>> I have these packages.
>>
>> drbd-pacemaker-8.4.2-1.fc17.i686
>> pacemaker-libs-1.1.7-2.fc17.i686
>> pacemaker-1.1.7-2.fc17.i686
>> corosync-2.3.0-1.fc17.i686
>> corosynclib-2.3.0-1.fc17.i686
>> pacemaker-cli-1.1.7-2.fc17.i686
>> pacemaker-cluster-libs-1.1.7-2.fc17.i686
>>
>> pacemaker-libs-1.1.9-0.1.70ad9fa.git.fc18.i686
>> pacemaker-1.1.9-0.1.70ad9fa.git.fc18.i686
>> drbd-pacemaker-8.4.2-1.fc18.i686
>> pacemaker-cluster-libs-1.1.9-0.1.70ad9fa.git.fc18.i686
>> pacemaker-cli-1.1.9-0.1.70ad9fa.git.fc18.i686
>> corosynclib-2.3.1-1.fc18.i686
>> corosync-2.3.1-1.fc18.i686
>>
>> The corosync config :
>>
>> totem {
>> version: 2
>> secauth: off
>> cluster_name: cluster
>>   interface {
>>         ringnumber: 0
>>         bindnetaddr: 192.168.1.0
>>         ttl: 1
>>   }
>> transport: udpu
>> }
>>
>> nodelist {
>>   node {
>>         ring0_addr: noeud1.xxxx.fr
>>        }
>>   node {
>>         ring0_addr: noeud2.xxxx.fr
>>        }
>> }
>>
>> quorum {
>> provider: corosync_votequorum
>> }
>>
>> logging {
>> to_syslog: yes
>> debug: off
>> }
>>
>> A short time after starting pacemaker I have this.
>>
>> FC18 node:
>>
>> Corosync Nodes:
>> noeud1.xxxx.fr noeud2.xxxx.fr
>> Pacemaker Nodes:
>> noeud1.xxxx.fr noeud1.xxxx.fr noeud2.xxxx.fr noeud2.xxxx.fr
>>
>> <node id="-33445696" uname="noeud2.xxxx.fr" type="normal"/>
>> <node id="-1062731267" uname="noeud1.xxxx.fr" type="normal"/>
>> <node id="3232236029" uname="noeud1.xxxx.fr"/>
>> <node id="4261521600" uname="noeud2.xxxx.fr"/>
>>
>> Why four nodes ?!? What are the nodes 3232236029 and 4261521600 ?
>
> The same as the other two but stored as %u (unsigned int) instead of %d (signed int).
> This was a bug in older versions, you can work around it by specifying a (small) nodeid in corosync.conf
>
>>
>> FC17 node:
>>
>> Corosync Nodes:
>> noeud1.xxxx.fr noeud2.xxxx.fr
>> Pacemaker Nodes:
>> noeud1.xxxx.fr noeud2.xxxx.fr
>>
>> <node id="-33445696" uname="noeud2.xxxx.fr" type="normal"/>
>> <node id="-1062731267" uname="noeud1.xxxx.fr" type="normal"/>
>>
>> On the FC17 I have some messages like this "error: cib_perform_op: Discarding update with feature set '3.0.7' greater than our own '3.0.6'".
>> On the FC18 "warning: cib_process_replace: Replacement 0.5.4 from noeud2.xxxx.fr not applied to 0.9.0: current epoch is greater than the replacement"
>>
>> Pacemaker 1.1.7 and 1.1.9 are not compatible ?
>
> This should provide some more information:
>     http://blog.clusterlabs.org/blog/2013/mixing-pacemaker-versions/
>
>