[pgpool-hackers: 261] Re: Road to 3.3

[Tatsuo Ishii]

Now that we have released 3.3 alpha, I revisited this to review for
the passed work toward 3.3.

From: Tatsuo Ishii <ishii at postgresql.org>
Subject: [pgpool-hackers: 171] Road to 3.3
Date: Tue, 26 Feb 2013 18:18:04 +0900 (JST)
Message-ID: <20130226.181804.308674087898613797.t-ishii at sraoss.co.jp>

> Hi pgpool-hackers,
> 
> I'm thinking about the road to pgpool-II 3.3. Here are
> additions/enhancements for 3.3 in my mind. Comments, suggestions are
> welcome.
> 
> 1) Easy installer
> 
> At least for a "simple configuration", it would be nice if we provide
> simple and easy installer. Here the "simple configuration" means
> two-node pgpool-II hot/standby configuration (probably PostgreSQL is
> on the same node). For this purpose we need the latest pgpool-II RPM
> (the latest RPM is pgpool-II 3.2.0 at this point as far as I know,
> while the latest version is pgpool-II 3.2.3) and simple installer
> written in shell. Native replication mode or streaming replication
> mode will be allowed.

Nozomi has been working on this and pretty close to finish. She is
going to push to git soon.

> 2) Enhance pgpoolAdmin
> 
> pgpoolAdmin should adopt the configuration above. i.e. pgpoolAdmin
> should be able to handle multiple pgpool-II instances.

Has been already pushed to git. Still some work remains though.

> 3) Import PostgreSQL 9.2 SQL parser
> 
> Current pgpool-II's parser is based on PostgreSQL 9.0. I would like to
> update it to 9.2 (Nozomi is already working on this BTW).

Done.

> 4) Various enhancements on watchdog
> 
> - Allow to watch pgpool-II even all PostgreSQL servers are
>   down. Currently if all PostgreSQL are down, watchdog treats
>   pgpool-II as down. Pgpool-II should concentrate on watching
>   PostgreSQL status and watchdog should concentrate on pgpool-II
>   status.

Done.

> - Prevent multiple pgpool-II instances from executing fail over script
>   at the same time.

Done.

> - Restart watchdog process if it accidentally dies

Done.

> - More secure protocol between watchdog

Done.

> 5) Misc
> 
> - Allow to share on memory query cache by multiple pgpool-II instances
>   using memcached. Currently multiple pgpool-II instances can share
>   one memcached server but has problem with cache invalidation.

Done.

> - Delete obsoleted (none on memory) query cache

Not yet. Probably due to limitation of developers time, it is likely
that is postponed to 3.4.

> - Add new directive to control the time out to look for new primary
>   node. Currently it is "recovery_timeout" which is too long(the
>   default is 90 seconds).

New developer Usama is working on this.

Remaining concers are mainly localized docs. Any volunteer for French
and Chinese docs?  If none, 3.3 will be shipped withput those
documentations.
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp