Re: Replication logging

Simon Riggs wrote:
> On Tue, 2011-01-18 at 10:57 -0500, Tom Lane wrote:
> > Magnus Hagander <magnus(at)hagander(dot)net> writes:
> > > Is there *any* usecase for setting them differently though?
> >
> > I can't believe we're still engaged in painting this bikeshed. Let's
> > just control it off log_connections and have done.
>
> Yes, this is a waste of time. Leave it as is because its there for a
> very good reason, as Robert already explained.
>
> The code was put in explicitly because debugging replication connections
> is quite important and prior to that addition, wasn't easy. It's a very
> separate thing from logging the hundreds/thousands of other connections
> on the system.
>
> I don't really care about neatness of code, or neatness of parameters. I
> want to be able to confirm the details of the connection.
> pg_stat_replication is dynamic, not a historical record of connections
> and reconnections.
>
> How else will you diagnose an erratic network, or an accidental change
> of authority? Replication is so important it isn't worth tidying away a
> couple of lines of log. My objective is to make replication work, not to
> reduce the size of the average log file by 1 or 2 lines.
>
> I'm particularly concerned that people make such changes too quickly.
> There are many things in this area of code that need changing, but also
> many more that do not. If we are to move forwards we need to avoid going
> one step forwards, one step back.

There were enough people who wanted a change that we went ahead and did
it --- if there was lack of agreement, we would have delayed longer.

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ It's impossible for everything to be true. +