Re: An attempt to avoid locally-committed-but-not-replicated-to-standby-transactions in synchronous replication

On Fri, Aug 5, 2022 at 8:19 AM Kyotaro Horiguchi
<horikyota(dot)ntt(at)gmail(dot)com> wrote:
>
> At Tue, 26 Apr 2022 08:26:59 +0200, Laurenz Albe <laurenz(dot)albe(at)cybertec(dot)at> wrote in
> > While this may mitigate the problem, I don't think it will deal with
> > all the cases which could cause a transaction to end up committed locally,
> > but not on the synchronous standby. I think that only using the full
> > power of two-phase commit can make this bulletproof.
> >
> > Is it worth adding additional complexity that is not a complete solution?
>
> I would agree to this. Likewise 2PC, whatever we do to make it
> perfect, we're left with unresolvable problems at least for now.
>
> Doesn't it meet your requirements if we have a means to know the last
> transaction on the current session is locally committed or aborted?
>
> We are already internally managing last committed LSN. I think we can
> do the same thing about transaction abort and last inserted LSN and we
> can expose them any way. This is way simpler than the (maybe)
> uncompletable attempt to fill up the deep gap.

There can be more txns that are
locally-committed-but-not-yet-replicated. Even if we have that
information stored somewhere, what do we do with it? Those txns are
committed from the client perspective but not committed from the
server's perspective.

Can you please explain more about your idea, I may be missing something?

--
Bharath Rupireddy
RDS Open Source Databases: https://aws.amazon.com/rds/postgresql/