[pgpool-general: 6095] Re: pgpool failover from node 1 to node 0: incorrect value for %P in version 3.7.3 ?

[Tatsuo Ishii]

Have a nice week-end too! I am looking forward to hearing from you
after coming back. Please don't hesitate to ask me regarding applying
the patch.

Best regards,
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese:http://www.sraoss.co.jp

> Hi Tatsuo
> 
> Thanks for the prompt response and for the patch ! I am glad you could reproduce it.
> 
> I am leaving now for a long week-end (monday is holiday in Belgium) and I am not familiar with patches to do it right now, but I'll definitively try when I am back. 
> 
> Have a good week-end
> Pierre 
> 
>     On Saturday, May 19, 2018, 3:50:28 AM GMT+2, Tatsuo Ishii <ishii at sraoss.co.jp> wrote:  
>  
>  Pierre,
> 
>> Hi,
>> I have a problem with pgpool 3.7.2 and pgpool 3.7.3, but not with pgpool 3.7.1 so it might be a bug.
>> When there is a failover from node 1 to node 0 then the parameter %P (old primary id) which is passed to the failover script is incorrect. It says 0 instead of 1. I encounter the error systematically when I do a first failover from node 0 to node 1, then a recovery of node 0, then a new failover from node 1 to node 0. In this new failover - from node 1 to node 0 - the parameter old_primary_id is not correctly passed.
>> I have reverted back to pgpool 3.7.1 and the problem has disappeared.
>> My config, for failover_command, is:
>> failover_command = '/scripts/failover.sh %d %h %p %D %m %H %M %P %r %R'                                   # Executes this command at failover                                   # Special values:                                   #   %d = node id                                   #   %h = host name                                   #   %p = port number                                   #   %D = database cluster path                                   #   %m = new master node id                                   #   %H = hostname of the new master node                                   #   %M = old master node id                                   #   %P = old primary node id                                   #   %r = new master port number
>> and in the log I see that the parameter %P is incorrect with pgpool 3.7.3 and pgpool 3.7.2 but that it is correct with pgpool 3.7.1
>> I can provide more logs and configuration if needed
>> Kind regards, 
>> Pierre 
> 
> Thanks for the report. I was able to reproduce the problem. I think
> problem with 3.7.2 and 3.7.3 is, they pass node id 0 if current
> primary node's status is down. Of course the status is down because
> it's the reason why the failover script gets called:-< 3.7.1 does not
> check the primary node status. So it correctly passes the failed
> primary node id.
> 
> I think the fix would be passes the primary node id regardless the
> status. Attached one line patch does the trick. At least it worked for
> me. So can you please try it out?
> 
> Best regards,
> --
> Tatsuo Ishii
> SRA OSS, Inc. Japan
> English: http://www.sraoss.co.jp/index_en.php
> Japanese:http://www.sraoss.co.jp