Horton

Re: row filtering for logical replication

Ajin Cherian <itsajin@gmail.com>

View thread
From: Ajin Cherian <itsajin@gmail.com>
To: Amit Kapila <amit.kapila16@gmail.com>
Cc: Dilip Kumar <dilipbalaut@gmail.com>, Euler Taveira <euler@eulerto.com>, Peter Smith <smithpb2250@gmail.com>, Rahila Syed <rahilasyed90@gmail.com>, Peter Eisentraut <peter.eisentraut@enterprisedb.com>, Önder Kalacı <onderkalaci@gmail.com>, japin <japinli@hotmail.com>, Michael Paquier <michael@paquier.xyz>, David Steele <david@pgmasters.net>, Craig Ringer <craig@2ndquadrant.com>, Tomas Vondra <tomas.vondra@2ndquadrant.com>, Amit Langote <amitlangote09@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2021-10-02T07:44:48Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Release cache tuple when no longer needed

  2. Add some additional tests for row filters in logical replication.

  3. Fix one of the tests introduced in commit 52e4f0cd47.

  4. Allow specifying row filters for logical replication of tables.

  5. Move scanint8() to numutils.c

  6. Replace Test::More plans with done_testing

  7. Reduce relcache access in WAL sender streaming logical changes

  8. Small cleanups related to PUBLICATION framework code

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

  10. Allow publishing the tables of schema.

  11. Doc: improve documentation of CREATE/ALTER SUBSCRIPTION.

  12. Add PublicationTable and PublicationRelInfo structs

  13. Remove unused argument "txn" in maybe_send_schema().

  14. Add prepare API support for streaming transactions in logical replication.

  15. Unify PostgresNode's new() and get_new_node() methods

  16. Use l*_node() family of functions where appropriate

  17. Add support for prepared transactions to built-in logical replication.

  18. Restore the portal-level snapshot after procedure COMMIT/ROLLBACK.

  19. Rename a parse node to be more general

  20. Remove unused column atttypmod from initial tablesync query

  21. SEARCH and CYCLE clauses

Attachments

On Wed, Sep 22, 2021 at 2:05 PM Ajin Cherian <itsajin@gmail.com> wrote:
>
> On Wed, Sep 22, 2021 at 1:50 PM Amit Kapila <amit.kapila16@gmail.com> wrote:
> >
> > On Wed, Sep 22, 2021 at 6:42 AM Ajin Cherian <itsajin@gmail.com> wrote:
> > >
> >
> > Why do you think that the second assumption (if there is an old tuple
> > it will contain all RI key fields.) is broken? It seems to me even
> > when we are planning to include unchanged toast as part of old_key, it
> > will contain all the key columns, isn't that true?
>
> Yes, I assumed wrongly. Just checked. What you say is correct.
>
> >
> > > I think we
> > > still need to deform both old tuple and new tuple, just to handle this case.
> > >
> >
> > Yeah, but we will anyway talking about saving that cost for later if
> > we decide to send that tuple. I think we can further try to optimize
> > it by first checking whether the new tuple has any toasted value, if
> > so then only we need this extra pass of deforming.
>
> Ok, I will go ahead with this approach.
>
> >
> > > There is currently logic in ReorderBufferToastReplace() which already
> > > deforms the new tuple
> > > to detoast changed toasted fields in the new tuple. I think if we can
> > > enhance this logic for our
> > > purpose, then we can avoid an extra deform of the new tuple.
> > > But I think you had earlier indicated that having untoasted unchanged
> > > values in  the new tuple
> > > can be bothersome.
> > >
> >
> > I think it will be too costly on the subscriber side during apply
> > because it will update all the unchanged toasted values which will
> > lead to extra writes both for WAL and data.
> >

Based on the discussion above, I've added two more slot pointers in
the RelationSyncEntry structure to store tuples that have been
deformed. Once the tuple (old and new) is deformed , then it is stored
in the structure, where it can be retrieved while writing to the
stream.I have also changed the logic so that the old tuple is not
populated, as Dilip pointed out, it will have all the RI columns if it
is changed.
I've added two new APIs in proto.c for writing tuple cached and
writing update cached. These are called if the the slots
contain previously deformed tuples.

I have for now also rebased the patch and merged the first 5 patches
into 1, and added my changes for the above into the second patch.

regards,
Ajin Cherian
Fujitsu Australia