Yoshiharu Mori a écrit :
>>>> Hello.
>>>>
>>>> I am using 3 nodes (2 slaves subscribed to one master).
>>>> Replication is fine, but FAILOVER command don't work correctly.
>>>>
>>>> I do FAILOVER from "node 1 (master)" to "node 3" as following.
>>>>
>>>>     failover ( id = 1, backup node = 3);
>>>>
>>>> I could confirm that node 3 was master and INSERT tupples 
>>>> updated to the slave (node 2) 
>>>>
>>>> However I couldn't FAILOVER from "node 1 (master)" to "node 2".
>>>>
>>>>     failover ( id = 1, backup node = 2);
>>>>
>>>> INSERT tupples didn't updated to the slave ( node 3).
>>>>
>>>>
>>>> I use postgresql-8.2.6 and slony-I 1.2.12 (Because of xxid.*.sql problems, I couldn't setup the slony-I-1.2.13--;).
>>>>
>>>> Any suggestion?
>>> I had checked the version diffrences.
>>>
>>> ---
>>> SUCCESS FAILOVER from node 1 to node 2
>>> ---
>>> slony-I 1.2.11
>>>
>>> ---
>>> FAIL  FAILOVER from node 1 to node 2
>>> ---
>>> slony-I 1.2.12
>>> slony-I 1.2.13
>>> slony-I 1.2.13 CVS-HEAD
>> How could it ever work ? AFAIK, you can't have different slony versions on
>> nodes involved in the same replication network.
> 
> Of cource,I use the same versions on nodes. 
> Only slony-I 1.2.11, I could confirm failover.
> 
> Regards,
> 

Oh, sorry, I misunderstood... You told us 1.2.13 has a buggy failover procedure?

Which could be related to problems Sébastien reported with failover.

Hey, I should test failover before updating to 1.2.13...

Regards,
SAS