Re: Proposal for Allow postgresql.conf values to be changed via SQL

Cédric Villemain <cedric@2ndquadrant.com>

From: Cédric Villemain <cedric@2ndquadrant.com>

To: Tom Lane <tgl@sss.pgh.pa.us>

Cc: Fujii Masao <masao.fujii@gmail.com>, Amit Kapila <amit.kapila@huawei.com>, pgsql-hackers@postgresql.org, Robert Haas <robertmhaas@gmail.com>, Greg Smith <greg@2ndquadrant.com>, Josh Berkus <josh@agliodbs.com>, Magnus Hagander <magnus@hagander.net>, Christopher Browne <cbbrowne@gmail.com>

Date: 2012-11-18T09:52:24Z

Lists: pgsql-hackers

Le samedi 17 novembre 2012 22:57:49, Tom Lane a écrit :
> Fujii Masao <masao.fujii@gmail.com> writes:
> > Do we really need to store the settings in a system table?
> > Since WAL would be generated when storing the settings
> > in a system table, this approach seems to prevent us from
> > changing the settings in the standby.
> 
> That's a really good point: if we try to move all GUCs into a system
> table, there's no way for a standby to have different values; and for
> some of them different values are *necessary*.
> 
> I think that shoots down this line of thought entirely.  Can we go
> back to the plain "write a file" approach now?  I think a "SET
> PERSISTENT" command that's disallowed in transaction blocks and just
> writes the file immediately is perfectly sensible.

I was justifying the usage of a table structure, not to keep it in sync (just 
use it to hide the complexity of locks).

Anyway that was just comments.
-- 
Cédric Villemain +33 (0)6 20 30 22 52
http://2ndQuadrant.fr/
PostgreSQL: Support 24x7 - Développement, Expertise et Formation