Horton

Re: row filtering for logical replication

Peter Smith <smithpb2250@gmail.com>

View thread
From: Peter Smith <smithpb2250@gmail.com>
To: Ajin Cherian <itsajin@gmail.com>
Cc: Greg Nancarrow <gregn4422@gmail.com>, Dilip Kumar <dilipbalaut@gmail.com>, Amit Kapila <amit.kapila16@gmail.com>, Euler Taveira <euler@eulerto.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: 2021-11-04T02:46:55Z
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

Hi.

During some ad-hoc filter testing I observed a quirk when there are
duplicate tables. I think we need to define/implement some proper
rules for this behaviour.

=====

BACKGROUND

When the same table appears multiple times in a CREATE PUBLICATION
then those duplicates are simply ignored. The end result is that the
table is only one time in the publication.

This is fine and makes no difference where there are no row-filters
(because the duplicates are all exactly the same as each other), but
if there *are* row-filters there there is a quirky behaviour.

=====

PROBLEM

Apparently it is the *first* of the occurrences that is used and all
the other duplicates are ignored.

In practice it looks like this.

ex.1)

DROP PUBLICATION
test_pub=# CREATE PUBLICATION p1 FOR TABLE t1 WHERE (a=1), t1 WHERE (a=2);
CREATE PUBLICATION
test_pub=# \dRp+ p1
                               Publication p1
  Owner   | All tables | Inserts | Updates | Deletes | Truncates | Via root
----------+------------+---------+---------+---------+-----------+----------
 postgres | f          | t       | t       | t       | t         | f
Tables:
    "public.t1" WHERE ((a = 1))

** Notice that the 2nd filter (a=2) was ignored

~

IMO ex1 is wrong behaviour. I think that any subsequent duplicate
table names should behave the same as if the CREATE was a combination
of CREATE PUBLICATION then ALTER PUBLICATION SET.

Like this:

ex.2)

test_pub=# CREATE PUBLICATION p1 FOR TABLE t1 WHERE (a=1);
CREATE PUBLICATION
test_pub=# ALTER PUBLICATION p1 SET TABLE t1 WHERE (a=2);
ALTER PUBLICATION
test_pub=# \dRp+ p1
                               Publication p1
  Owner   | All tables | Inserts | Updates | Deletes | Truncates | Via root
----------+------------+---------+---------+---------+-----------+----------
 postgres | f          | t       | t       | t       | t         | f
Tables:
    "public.t1" WHERE ((a = 2))

** Notice that the 2nd filter (a=2) overwrites the 1st filter (a=1) as expected.

~~

The current behaviour of duplicates becomes even more "unexpected" if
duplicate tables occur in a single ALTER PUBLICATION ... SET command.

ex.3)

test_pub=# CREATE PUBLICATION p1;
CREATE PUBLICATION
test_pub=# ALTER PUBLICATION p1 SET TABLE t1 WHERE (a=1), t1 WHERE (a=2);
ALTER PUBLICATION
test_pub=# \dRp+ p1
                               Publication p1
  Owner   | All tables | Inserts | Updates | Deletes | Truncates | Via root
----------+------------+---------+---------+---------+-----------+----------
 postgres | f          | t       | t       | t       | t         | f
Tables:
    "public.t1" WHERE ((a = 1))


** Notice the 2nd filter (a=2) did not overwrite the 1st filter (a=1).
I think a user would be quite surprised by this behaviour.

=====

PROPOSAL

I propose that we change the way duplicate tables are processed to
make it so that it is always the *last* one that takes effect (instead
of the *first* one). AFAIK doing this won't affect any current PG
behaviour, but doing this will let the new row-filter feature work in
a consistent/predictable/sane way.

Thoughts?

------
Kind Regards,
Peter Smith.
Fujitsu Australia