Re: Multi-Master Logical Replication

On Fri, Jun 3, 2022 at 7:12 AM Bruce Momjian <bruce(at)momjian(dot)us> wrote:
>
> On Thu, Jun 2, 2022 at 05:12:49PM +1000, Peter Smith wrote:
> > On Thu, Jun 2, 2022 at 12:03 AM Bruce Momjian <bruce(at)momjian(dot)us> wrote:
> > >
> > > On Wed, Jun 1, 2022 at 10:27:27AM +0530, Amit Kapila wrote:
> > ...
> >
> > > My big point is that you should not be showing up with a patch but
> > > rather have these discussions to get agreement that this is the
> > > direction the community wants to go.
> >
> > The purpose of posting the POC patch was certainly not to present a
> > fait accompli design/implementation.
> >
> > We wanted to solicit some community feedback about the desirability of
> > the feature, but because LRG is complicated to describe we felt that
> > having a basic functional POC might help to better understand the
> > proposal. Also, we thought the ability to experiment with the proposed
> > API could help people to decide whether LRG is something worth
> > pursuing or not.
>
> I don't think the POC is helping, and I am not sure we really want to
> support this style of architecture due to its complexity vs other
> options.
>

None of the other options discussed on this thread appears to be
better or can serve the intent. What other options do you have in mind
and how are they simpler than this? As far as I can understand this
provides a simple way to set up n-way replication among nodes.

I see that other databases provide similar ways to set up n-way
replication. See [1] and in particular [2][3][4] provides a way to set
up n-way replication via APIs. Yet, another way is via configuration
as seems to be provided by MySQL [5] (Group Replication Settings).
Most of the advantages have already been shared but let me summarize
again the benefits it brings (a) more localized database access for
geographically distributed databases, (b) ensuring continuous
availability in case of the primary site becomes unavailable due to a
system or network outage, any natural disaster on the site, (c)
environments that require a fluid replication infrastructure, where
the number of servers has to grow or shrink dynamically and with as
few side-effects as possible. For instance, database services for the
cloud, and (d) load balancing. Some of these can probably be served in
other ways but not everything.

I see your point about POC not helping here and it can also sometimes
discourage OP if we decide not to do this feature or do it in an
entirely different way. But OTOH, I don't see it stopping us from
discussing the desirability or design of this feature.

[1] - https://docs.oracle.com/cd/E18283_01/server.112/e10707/rarrcatpac.htm
[2] - https://docs.oracle.com/cd/E18283_01/server.112/e10707/rarrcatpac.htm#i96251
[3] - https://docs.oracle.com/cd/E18283_01/server.112/e10707/rarrcatpac.htm#i94500
[4] - https://docs.oracle.com/cd/E18283_01/server.112/e10707/rarrcatpac.htm#i97185
[5] - https://dev.mysql.com/doc/refman/8.0/en/group-replication-configuring-instances.html

--
With Regards,
Amit Kapila.