Horton

Re: Skipping logical replication transactions on subscriber side

Amit Kapila <amit.kapila16@gmail.com>

View thread View on postgr.es
From: Amit Kapila <amit.kapila16@gmail.com>
To: Masahiko Sawada <sawada.mshk@gmail.com>
Cc: Greg Nancarrow <gregn4422@gmail.com>, Peter Eisentraut <peter.eisentraut@enterprisedb.com>, "houzj.fnst@fujitsu.com" <houzj.fnst@fujitsu.com>, "tanghy.fnst@fujitsu.com" <tanghy.fnst@fujitsu.com>, "osumi.takamichi@fujitsu.com" <osumi.takamichi@fujitsu.com>, Alexey Lesovsky <lesovsky@gmail.com>, PostgreSQL-development <pgsql-hackers@postgresql.org>
Date: 2021-10-28T04:05:15Z
Lists: pgsql-hackers 
On Thu, Oct 28, 2021 at 7:49 AM Masahiko Sawada <sawada.mshk@gmail.com> wrote:
>
> On Wed, Oct 27, 2021 at 2:43 PM Amit Kapila <amit.kapila16@gmail.com> wrote:
> >
> > On Wed, Oct 27, 2021 at 10:43 AM Masahiko Sawada <sawada.mshk@gmail.com> wrote:
> > >
> > > > > BTW how useful is specifying LSN instead of XID in practice? Given
> > > > > that this skipping behavior is used to skip the particular transaction
> > > > > (or its part of operations) in question, I’m not sure specifying LSN
> > > > > or time is useful.
> > > > >
> > > >
> > > > I think if the user wants to skip multiple xacts, she might want to
> > > > use the highest LSN to skip instead of specifying individual xids.
> > >
> > > I think it assumes that the situation where the user already knows
> > > multiple transactions that cannot be applied on the subscription but
> > > how do they know?
> > >
> >
> > Either from the error messages in the server log or from the new view
> > we are planning to add. I think such a case is possible during the
> > initial synchronization phase where apply worker went ahead then
> > tablesync worker by skipping to apply the changes on the corresponding
> > table. After that it is possible, that table sync worker failed during
> > copy and apply worker fails during the processing of some other rel.
>
> Does it mean that if both initial copy for the corresponding table by
> table sync worker and applying changes for other rels by apply worker
> fail, we skip both by specifying LSN?
>

Yes.

> If so, can't we disable the
> initial copy for the table and skip only the changes for other rels
> that cannot be applied?
>

But anyway you need some way to skip changes via a particular
tablesync worker so that it can mark the relation in 'ready' state.  I
think one can also try to use disable_on_error option in such
scenarios depending on how we expose it. Say, if the option means that
all workers (apply or table sync) should be disabled on an error then
it would be a bit tricky but if we can come up with a way to behave
differently for different workers then it is possible to disable one
set of workers and skip the changes in another set of workers.

-- 
With Regards,
Amit Kapila.

Commits

  1. Test ALIGNOF_DOUBLE==4 compatibility under ALIGNOF_DOUBLE==8.

  2. Reorder subskiplsn in pg_subscription to avoid alignment issues.

  3. Add ALTER SUBSCRIPTION ... SKIP.

  4. Optionally disable subscriptions on error.

  5. Update docs of logical replication for commit 8d74fc96db.

  6. Respect permissions within logical replication.

  7. Fix regression test failure caused by commit 8d74fc96db.

  8. Add a view to show the stats of subscription workers.

  9. Add logical change details to logical replication worker errcontext.

  10. Rename LOGICAL_REP_MSG_STREAM_END to LOGICAL_REP_MSG_STREAM_STOP.

  11. Fix typo in protocol.sgml.

  12. Remove unused argument in apply_handle_commit_internal().

  13. Fix replication of in-progress transactions in tablesync worker.

  14. Reorder pg_sequence columns to avoid alignment issue