Re: Is a VACUUM or ANALYZE necessary after logical replication?

From: Achilleas Mantzios <a(dot)mantzios(at)cloud(dot)gatewaynet(dot)com>
To: pgsql-general(at)lists(dot)postgresql(dot)org
Subject: Re: Is a VACUUM or ANALYZE necessary after logical replication?
Date: 2024-06-16 06:54:58
Message-ID: f22ed53a-8ebb-45e9-98df-7f5506b3ae1c@cloud.gatewaynet.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

Στις 16/6/24 02:13, ο/η Adrian Klaver έγραψε:
> On 6/15/24 15:55, Koen De Groote wrote:
>> I've gone over all of
>> https://www.postgresql.org/docs/current/logical-replication.html
>> <https://www.postgresql.org/docs/current/logical-replication.html>
>> and the only mentions of the word "index" I could find was in
>> relation to replica identity and examples of table definitions
>> showing primary key indexes.
>>
>> Nothing is said about indexes. Maybe for good reason, maybe they are
>> fully functionality immediately after replication?
>>
>> So the main question: Once a table is fully replicated, do I need to
>> vacuum(analyze) that table, or are the indexes on that table already
>> functional?
>
> VACUUM/ANALYZE is not about making the index functional. The VACUUM
> marks the space dead tuples occupy in the table and associated indexes
> as available for recycling. The ANALYZE updates tables statistics to
> help the planner make decisions on what query plan to use. On a fresh
> table VACUUM will not be of much value, ANALYZE though will help by
> creating up to date table statistics.

Hi Adrian, however in case the replication is problematic due to e.g. :

- wrong encoding to the new system, e.g. from SQL_ASCII to UTF-8

- ALWAYS triggers written without full schema qualification or other
problems throwing ERRORs, etc

- server restarts during the sync phase

- etc

those will produce rollbacks, hence bloating right from the start. In
those cases either VACUUM FULL will be needed, or even better correct
the errors in their source and repeat the whole process. Normally the
new DB (subscriber) should be a little smaller than the original
(publisher).

As you said, in any case, ANALYZE will be always needed afterwards, as
well as taking care of sequences. IMHO the sequence part would be nice
to be handled in a more elegant manner, e.g. by an option in pg_dump to
dump only sequences, for "nearly zero" downtime upgrades, this step
should happen rapidly at the switchover.

>
>>
>> Regards,
>> Koen De Groote
>
--
Achilleas Mantzios
IT DEV - HEAD
IT DEPT
Dynacom Tankers Mgmt (as agents only)

In response to

Browse pgsql-general by date

  From Date Subject
Next Message David G. Johnston 2024-06-16 07:09:17 Re: Is a VACUUM or ANALYZE necessary after logical replication?
Previous Message Laurenz Albe 2024-06-16 04:53:31 Re: DROP COLLATION vs pg_collation question