Brad Nicholson wrote:
> On Fri, 2006-09-29 at 09:55 -0700, Jeff Davis wrote:
>   
>> 	
>> The Slony-1.2 release notes say:
>>
>> "The new behaviour points more towards
>>   "init" / "rc.d"-like handling, where, upon bootup, one "rc.d" script
>>   might start up PostgreSQL, another one starts pgpool, and a third
>>   (which must run third!) starts up a slon."
>>
>> Why must a slon start after pgpool?
>>
>> Regards,
>> 	Jeff Davis
>>     
>
> I assume the slon is going to connect through pgpool - it can't do that
> if the pool hasn't started.
>   
That strikes me as a terrible idea (e.g. - running slon thru pgpool);
the slon really needs to talk to the RIGHT node, and pgpool can make
that somewhat nondeterministic.

Hmm.  [rummaging thru CVS logs...]  I committed that change...

I must have been a bit off when I wrote that; starting up the slon
should be quite independent of starting up pgpool.

The only reason to want pgpool to already be running would be if it were
greatly more Slony-I-aware, such that you might submit requests to
pgpool that would cause reconfiguration of a Slony-I cluster.  That
would point more to needing the slons running before starting pgpool,
rather than the reverse.  And there's *not* such deep/tight integration
between pgpool and Slony-I.

I'm revising that comment...