[pgpool-hackers: 850] Re: Small problem in pgpool-II side implementation of pcp_detach_node

[Tatsuo Ishii]

Usama,

Good catch! I was surprised that this had not been reported until
today. Definitely the behavior of pcp_detach_node gracefully is broken
and should be fixed as you described. Could you please go ahead and
fix it?

Also I think is another reason we release minor version up.

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

> Hi
> 
> I think there is a problem in pgpool-II side implementation of
> pcp_detach_node command. pcp_detach_node should check if it is allowed to
> process detach_node command on the particular node before blocking the
> incoming connections and closing the existing connections.
> 
> This is the problem in the case when gracefully detaching of the node is
> requested.
> 
> Consider the example, If the pgpool-II is connected to 2 backend nodes.
> And user asks the pgpool-II to detach node-id 3
> 
> usama at ubuntu: pcp_detach_node -U postgres -g -d 3
> 
> Now, as node id 3 is not a valid backend node in this situation, The
> command should immediately return with an error instead it disallows the
> new incoming connections, close the active connection only to find out the
> operation is not allowed.
> 
> What are your thoughts on it? Should I go ahead and make this above change?
> 
> Thanks
> Best regards
> Muhammad Usama