Re: Added schema level support for publication.

From: vignesh C <vignesh21(at)gmail(dot)com>
To: Bharath Rupireddy <bharath(dot)rupireddyforpostgres(at)gmail(dot)com>
Cc: Amit Kapila <amit(dot)kapila16(at)gmail(dot)com>, PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>
Subject: Re: Added schema level support for publication.
Date: 2021-01-10 17:51:16
Message-ID: CALDaNm1wOma56yeOFf_LxHeb9aAr1oY1WPtYHXfBLw=O=3NsLg@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Thanks for your comments Bharath, please find my opinion below.

On Sat, Jan 9, 2021 at 8:08 PM Bharath Rupireddy
<bharath(dot)rupireddyforpostgres(at)gmail(dot)com> wrote:
> I think this feature can be useful, in case a user has a lot of tables
> to publish inside a schema. Having said that, I wonder if this feature
> mandates users to create the same schema with same
> permissions/authorizations manually on the subscriber, because logical
> replication doesn't propagate any ddl's so are the schema or schema
> changes? Or is it that the list of tables from the publisher can go
> into a different schema on the subscriber?
>

DDL's will not be propagated to the subscriber. Users have to create
the schema & tables in the subscriber. No change in
Permissions/authorizations handling, it will be the same as the
existing behavior for relations.

> Since the schema can have other objects such as data types, functions,
> operators, I'm sure with your feature, non-table objects will be
> skipped.
>

Yes, only table data will be sent to subscribers, non-table objects
will be skipped.

> As Amit pointed out earlier, the behaviour when schema dropped, I
> think we should also consider when schema is altered, say altered to a
> different name, maybe we should change that in the publication too.
>

I agree that when schema is altered the renamed schema should be
reflected in the publication.

> In general, what happens if we have some temporary tables or foreign
> tables inside the schema, will they be allowed to send the data to
> subscribers?
>

Temporary tables & foreign tables will not be added to the publications.

> And, with this feature, since there can be many huge tables inside a
> schema, the initial table sync phase of the replication can take a
> while.
>

Yes this is required.

> Say a user has created a publication for a schema with hundreds of
> tables in it, at some point later, can he stop replicating a single or
> some tables from that schema?
>

There is no provision for this currently.

> IMO, it's better to have the syntax - CREATE PUBLICATION
> production_publication FOR ALL TABLES IN SCHEMA production - just
> added IN between for all tables and schema.
>

I'm ok with the proposed syntax, I would like others' opinion too
before making the change.

> Say a user has a schema with 121 tables in it, and wants to replicate
> only 120 or 199 or even lesser tables out of it, so can we have some
> skip option to the new syntax, something like below?
> CREATE PUBLICATION production_publication FOR ALL TABLES SCHEMA
> production WITH skip = marketing, accounts, sales; --> meaning is,
> replicate all the tables in the schema production except marketing,
> accounts, sales tables.
>

Yes this is a good use case, will include this change.

Thanks for the comments, I will handle the comments and post a patch for this.

Regards,
Vignesh
EnterpriseDB: http://www.enterprisedb.com

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message vignesh C 2021-01-10 17:52:39 Re: Added schema level support for publication.
Previous Message Justin Pryzby 2021-01-10 17:29:45 Re: [HACKERS] Custom compression methods