[Pacemaker] Unable to bring up cluster after crash and update

[Andrew Beekhof]

Have you had a chance to try with Steve's latest corosync changes?
He thinks it should now be working.

On Fri, Jul 24, 2009 at 6:39 PM, Jonathan<jdccdevel at gmail.com> wrote:
> Sorry to reply to myself, but I wanted to update
>
> I modified my corosync.conf to set nodeid manually, and I am getting the
> same error.
>
> I have noticed something interesting in the debug logs, which may
> provide an indication of what is going on:
>
> The host name is Aries, and the nodeid in the config is 9001001
>
> Just before a bunch of forking happens we have the following:
>
> Jul 24 10:20:07 corosync [pcmk  ] plugin.c:419 info: pcmk_plugin_init:
> Local node id: 0
> Jul 24 10:20:07 corosync [pcmk  ] plugin.c:420 info: pcmk_plugin_init:
> Local hostname: Aries
> Jul 24 10:20:07 corosync [pcmk  ] utils.c:234 info: update_member:
> Creating entry for node 0 born on 0
> Jul 24 10:20:07 corosync [pcmk  ] utils.c:261 info: update_member:
> 0x818be68 Node 0 now known as Aries (was: (null))
> Jul 24 10:20:07 corosync [pcmk  ] utils.c:277 info: update_member: Node
> Aries now has 1 quorum votes (was 0)
> Jul 24 10:20:07 corosync [pcmk  ] utils.c:287 info: update_member: Node
> 0/Aries is now: member
>
> Then, just before the segfault, we have:
>
> Jul 24 10:20:07 corosync [pcmk  ] plugin.c:633 notice: pcmk_peer_update:
> Stable membership event on ring 780: memb=1, new=1, lost=0
> Jul 24 10:20:07 corosync [pcmk  ] utils.c:234 info: update_member:
> Creating entry for node 9001001 born on 780
> Jul 24 10:20:07 corosync [pcmk  ] utils.c:287 info: update_member: Node
> 9001001/unknown is now: member
> Jul 24 10:20:07 corosync [pcmk  ] plugin.c:661 info: pcmk_peer_update:
> NEW:  .pending. 9001001
> Jul 24 10:20:07 corosync [pcmk  ] plugin.c:667 debug: pcmk_peer_update:
> Node 9001001 has address r(0) ip(172.29.1.1)
> Jul 24 10:20:07 corosync [pcmk  ] plugin.c:679 info: pcmk_peer_update:
> MEMB: .pending. 9001001
> Jul 24 10:20:07 corosync [pcmk  ] plugin.c:596 info:
> ais_mark_unseen_peer_dead: Node Aries was not seen in the previous
> transition
> Jul 24 10:20:07 corosync [pcmk  ] utils.c:287 info: update_member: Node
> 0/Aries is now: lost
> Jul 24 10:20:07 corosync [pcmk  ] plugin.c:712 debug: pcmk_peer_update:
> 2 nodes changed
> Jul 24 10:20:07 corosync [pcmk  ] plugin.c:1187 info:
> send_member_notification: Sending membership update 780 to 0 children
> Jul 24 10:20:07 corosync [pcmk  ] plugin.c:1439 CRIT: send_cluster_id:
> Assertion failure line 1439: local_nodeid != 0
>
> This is a change event for 2 nodes, Node 0/Aries and 9001001/unknown
>
> There is only one node in the cluster at this point, so shouldn't the
> node be 9001001/Aries ?
>
>
> Jonathan wrote:
>> I am not setting nodeid: in corosync.conf
>>
>> My understanding is that, in that case, the nodeid should be automaticly
>> generated from the ip address of ring0
>> Is that not the case?
>>
>> I have included a copy of my corosync.conf
>>
>> The only difference between the corosync.conf on the different nodes is
>> the IP address.
>>
>> Jonathan
>>
>> Andrew Beekhof wrote:
>>
>>> On Thu, Jul 23, 2009 at 5:36 AM, Jonathan deBoer<jono.deboer at gmail.com> wrote:
>>>
>>> I replied to this on the openais list
>>>
>>>
>>>
>>>> Jul 22 21:18:22 corosync [pcmk  ] plugin.c:1439 CRIT: send_cluster_id: Assertion failure line 1439: local_nodeid != 0
>>>>
>>>>
>>> What value of nodeid: is set in corosync.conf?
>>>
>>> _______________________________________________
>>> Pacemaker mailing list
>>> Pacemaker at oss.clusterlabs.org
>>> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>>
>>>
>>>
>>
>>
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> Pacemaker mailing list
>> Pacemaker at oss.clusterlabs.org
>> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>
>
>
> --
> J. deBoer Computer Consulting.
> 42 Birchmont Dr.
> Leduc, AB.
> T9E-8S4
> cell: 780-717-0669
>
>
> _______________________________________________
> Pacemaker mailing list
> Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>