Re: pg_config, pg_service.conf, postgresql.conf ....

Bruce Momjian wrote:
> Lamar Owen wrote:
> 
>>On Monday 27 February 2006 21:09, Bruce Momjian wrote:
>>
>>>One question I have is how this feature would be an improvement over
>>>just pointing pg_ctl at a postgresql.conf configuration file.  That
>>>config file has the ability to specify most if not all server
>>>parameters.
>>
>>The big problem is that postgresql.conf is dynamically generated during 
>>initdb, and its location depends upon initdb's parameters directly.  This 
>>makes it difficult to distribute, at least for packagers, a template of 
>>postgresql.conf or a 'default' postgresql.conf that plays nice with multiple 
>>versions and clusters, yet has centralized database tracking.
> 
> 
> But looking at postgresql.conf I see:
> 
>     #data_directory = 'ConfigDir'           # use data in another directory
>     ...
>     #port = 5432
> 
> so it seems everything in this configuration file is going to be
> duplicated in postgresql.conf.
> 
> We are adding an "include" capability for postgresql.conf.  Does that help?
> 
> Also, keep in mind this TODO item:
>     
>     * Allow pg_ctl to work properly with configuration files located outside
>       the PGDATA directory
>     
>       pg_ctl can not read the pid file because it isn't located in the
>       config directory but in the PGDATA directory.  The solution is to
>       allow pg_ctl to read and understand postgresql.conf to find the
>       data_directory value.
> 
> I am thinking it should be fixed as part of this.
> 
> What if we add an option to initdb to allow the user to specify the name
> and location of the postgresql.conf file?
> 

That is certainly a way to approach it, I see the tough bit being the 
parsing of postgresql.conf to figure out which parts of the global 
included file to ignore (i.e the stuff for the *other* clusters).

Would this work for the situation where you have older clusters on the 
box (versions that don't understand 'include')?

Additionally this would need to tackle start|stop etc for all clusters...

Cheers

Mark