[Pgpool-hackers] New patches for pcp_promote_node

[Gilles Darold]

Le 04/03/2011 01:50, Tatsuo Ishii a écrit :
> Thanks.
>
> BTW, I have a question: in my understanding, what pcp_promote_node
> will do is a) create a protomoe request in shared memory area, b) do a
> failover and restart child process. Promoting standby node to primary
> seems out of scope of this tool.
>
> We can do this by 1)stop PostgreSQL primary node, 2) then create a
> trigger file on the appropreate standby node 3) thsese steps can be
> managed by fail over scrip of pgpool-II. So it seems we can do what
> you want in current pgpool without using pcp_protomote_node.
>
> WHat's the use case for pcp_promoto_node?
Yes, it is almost possible to do a lot of thing manually or by scripting
with pgpool. This patch I think is useful because you have a command to
tell to pgpool who should be the new master and let pgpool doing the
right things internally. In this patch actually it don't do much more
than you explain above but in the future we can take fine control of
this step, like pausing client connections during the switch or anything
else.

-- 
Gilles Darold
http://dalibo.com - http://dalibo.org