On Monday 30 November 2009 17:46:45 Hans-Jürgen Schönig wrote:
> On Nov 30, 2009, at 10:32 AM, Stefan Kaltenbrunner wrote:
> > the question is if filtering on the sending side is actually the
> > "right thing" to do.
> > It increases the overhead and the complexity on the master,
> > especially if you think about different (partial) replication
> > agreements for different slaves and it might also be hard to
> > integrate with the planned sync/async modes.
> > On the other hand if you filter on the master you might be able to
> > avoid a lot of network traffic du to filtered wal records.
> > I think for a first step it might make more sense to look into doing
> > the filtering on the receiving side and look into actual integration
> > with SR at a later stage.
> one problem with not-filtering on the master is that you will end up
> with a lot of complexity if you start adding new tables to a replica
> because you just cannot add tables as easy as when you are doing stuff
> on the slave. the procedure seems ways more complex.
> in addition to that you are sending WAL which has to be discarded
> anyway.
> we thought about filtering "outside the master" a lot but to me it did
> not sound like good plan.
One possibility for the far future would be to allow filtering on a slave as
well:
master ---- full replication ---> primary slave --- split ---> slaves
Possibly doing only catalog recovery on the primary slave. In my opinion thats
heaps more complex and not better in all situation. So I would probably write
it down as a nice idea but not more.
Andres
