Horton

Re: row filtering for logical replication

Ajin Cherian <itsajin@gmail.com>

View thread
From: Ajin Cherian <itsajin@gmail.com>
To: Euler Taveira <euler@eulerto.com>
Cc: Peter Smith <smithpb2250@gmail.com>, Amit Kapila <amit.kapila16@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: 2021-09-20T09:47:43Z
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 8, 2021 at 7:59 PM Ajin Cherian <itsajin@gmail.com> wrote:
>
> On Wed, Sep 1, 2021 at 9:23 PM Euler Taveira <euler@eulerto.com> wrote:
> >

> Somehow this approach of either new_tuple or old_tuple doesn't seem to
> be very fruitful if the user requires that his replica is up-to-date
> based on the filter condition. For that, I think you will need to
> convert UPDATES to either INSERTS or DELETES if only new_tuple or
> old_tuple matches the filter condition but not both matches the filter
> condition.
>
> UPDATE
> old-row (match)       new-row (no match)  -> DELETE
> old-row (no match)  new row (match)       -> INSERT
> old-row (match)       new row (match)       -> UPDATE
> old-row (no match)  new-row (no match)  -> (drop change)
>

Adding a patch that strives to do the logic that I described above.
For updates, the row filter is applied on both old_tuple
and new_tuple. This patch assumes that the row filter only uses
columns that are part of the REPLICA IDENTITY. (the current patch-set
only
restricts this for row-filters that are delete only)
The old_tuple only has columns that are part of the old_tuple and have
been changed, which is a problem while applying the row-filter. Since
unchanged REPLICA IDENTITY columns
are not present in the old_tuple, this patch creates a temporary
old_tuple by getting such column values from the new_tuple and then
applies the filter on this hand-created temp old_tuple. The way the
old_tuple is created can be better optimised in future versions.

This patch also handles the problem reported by Houz in [1]. The patch
assumes a fix proposed by Dilip in [2]. This is the case
where toasted unchanged RI columns are not detoasted in the new_tuple
and has to be retrieved from disk during decoding. Dilip's fix
involved updating the detoasted value in the old_tuple when writing to
WAL. In the problem reported by Hou, when the row filter
is applied on the new_tuple and the decoder attempts to detoast the
value in the new_tuple and if the table was deleted at that time, the
decode fails.
To avoid this, in such a situation, the untoasted value in the
old_tuple (fix by Dilip) is copied to the new_tuple before the
row_filter is applied.
I have also refactored the way Peter initializes the row_filter by
moving it into a separate function before the insert/update/delete
specific logic is applied.

I have not changed any of the first 5 patches, just added my patch 006
at the end. Do let me know of any comments on this approach.

[1] - https://www.postgresql.org/message-id/OS0PR01MB571618736E7E79309A723BBE94E99%40OS0PR01MB5716.jpnprd01.prod.outlook.com
[2] - https://www.postgresql.org/message-id/OS0PR01MB611342D0A92D4F4BF26C0F47FB229@OS0PR01MB6113.jpnprd01.prod.outlook.com
regards,
Ajin Cherian
Fujitsu Australia