[Pacemaker] Upgrading from 1.0 to 1.1

[Kiril Proskurin]

29 июля 2011, 04:34 от Andrew Beekhof <andrew at beekhof.net>:






On Wed, Jul 27, 2011 at 5:02 PM, Proskurin Kirill
<
k.proskurin at corp.mail.ru> wrote:
> 27.07.2011 5:56, Andrew Beekhof пишет:
>>
>> On Tue, Jul 19, 2011 at 5:40 PM, Proskurin Kirill
>> <
k.proskurin at corp.mail.ru>  wrote:
>>>
>>> On 07/19/2011 03:22 AM, Andrew Beekhof wrote:
>>>>
>>>> On Fri, Jul 15, 2011 at 10:33 PM, Proskurin Kirill
>>>> <
k.proskurin at corp.mail.ru>    wrote:
>>>>>
>>>>> Hello all.
>>>>>
>>>>> I found what I using corosync with pacemaker "ver:0" with installed
>>>>> pacemaker 1.1.5 - eg without start a pacemakerd.
>>>>>
>>>>> Sounds wrong. :-)
>>>>> So I try to upgrade.
>>>>> I shutdown one node. Change 0 to 1 on service.d/pcmk
>>>>> Start corosync and then start pacemakerd via init script.
>>>>>
>>>>> But this node stays online and on clusters DC I see:
>>>>> cib: [18392]: WARN: cib_peer_callback: Discarding cib_sync_one message
>>>>> (255)
>>>>> from mysender10.example.com: not in our membership
>>>>
>>>> Thats odd.  The only you changed was ver: 0 to ver: 1 ?
>>>
>>> Yes, only this. To make it more clear - I have 4 nodes with ver 0 and try
>>> to
>>> add one with ver 1 and got this.
>>>
>>> Well I shutdown all nodes change all to 1 and star them up add all was
>>> ok.
>>> Not a really good way to upgrade but I don`t have time.
>>
>> Do you still have the logs for the failure case?
>> I'd really like to see them.
>
> No I don`t. But some time ago I got same error on vise-versa situation -
> then I try to add node with "ver: 0" to cluster there all nodes are "ver: 1"
>
> Anyway my cluster are down now so I can do some test. I will sent logs to
> maillist if I reproduce this situation again.

excellent

I already try it with 3 node test cluster.


I try 2 nodes with ver:1 and one with ver:0 - ver:0 stays offline. 
And i try vise-versa - 2 with ver:0 and 1 with ver:1 - ver:1 stays offline
I post this on #linix-cluster IRC and check it with danfrincu. 



Hm, ok I have some time before production start and make test again to send you logs. 
  

--  
Best regards, 
Proskurin Kirill
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://oss.clusterlabs.org/pipermail/pacemaker/attachments/20110729/c295d949/attachment.html>