Horton

Re: row filtering for logical replication

Euler Taveira <euler@eulerto.com>

View thread
From: "Euler Taveira" <euler@eulerto.com>
To: "Peter Smith" <smithpb2250@gmail.com>, "Amit Kapila" <amit.kapila16@gmail.com>
Cc: "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: 2021-08-25T00:21:57Z
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

On Tue, Aug 24, 2021, at 4:46 AM, Peter Smith wrote:
> I have used debug logging to confirm that what Amit wrote [1] is
> correct; the row-filter ExprState of *every* table's row_filter will
> be invalidated (and so subsequently gets rebuilt) when the user
> changes the PUBLICATION tables. This was a side-effect of the
> rel_sync_cache_publication_cb which is freeing the cached ExprState
> and setting the entry->replicate_valid = false; for *every* entry.
> 
> So yes, the ExprCache is getting rebuilt for some situations where it
> is not strictly necessary to do so.
I'm afraid we are overenginnering this feature. We already have a cache
mechanism that was suggested (that shows a small improvement). As you said the
gain for this new improvement is zero or minimal (it depends on your logical
replication setup/maintenance).

> 1. Although the ExprState cache is effective, in practice the
> performance improvement was not very much. My previous results [2]
> showed only about 2sec saving for 100K calls to the
> pgoutput_row_filter function. So I think eliminating just one or two
> unnecessary calls in the get_rel_sync_entry is going to make zero
> observable difference.
> 
> 2. IMO it is safe to expect that the ALTER PUBLICATION is a rare
> operation relative to the number of times that pgoutput_row_filter
> will be called (the pgoutput_row_filter is quite a "hot" function
> since it is called for every INSERT/UPDATE/DELETE). It will be orders
> of magnitude difference 1:1000, 1:100000 etc.
> 
> ~~
> 
> Anyway, I have implemented the suggested cache change because I agree
> it is probably theoretically superior, even if in practice there is
> almost no difference.
I didn't inspect your patch carefully but it seems you add another List to
control this new cache mechanism. I don't like it. IMO if we can use the data
structures that we have now, let's implement your idea; otherwise, -1 for this
new micro optimization.

[By the way, it took some time to extract what you changed. Since we're trading
patches, I personally appreciate if you can send a patch on the top of the
current one. I have some changes too and it is time consuming incorporating
changes in the main patch.]


--
Euler Taveira
EDB   https://www.enterprisedb.com/