Re: pgsql 8.3 : force FK (and consequently ON DELETE CASCADE) to be run under session_replication_role TO 'replica'

From: Peter Eisentraut <peter(dot)eisentraut(at)2ndquadrant(dot)com>
To: Achilleas Mantzios <achill(at)matrix(dot)gatewaynet(dot)com>, pgsql-admin(at)lists(dot)postgresql(dot)org
Subject: Re: pgsql 8.3 : force FK (and consequently ON DELETE CASCADE) to be run under session_replication_role TO 'replica'
Date: 2019-07-05 16:42:40
Message-ID: f0d2cbbb-ac4f-4dc0-7986-973fd1b311b4@2ndquadrant.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-admin

On 2019-07-05 12:17, Achilleas Mantzios wrote:
> we run some replication statements code under with session_replication_role TO 'replica' ,
> however this disables FK constraints and consequently some important ON DELETE CASCADE that must be run.
>
> I tested this and this works :
> psql -qt -c 'SELECT '\''ALTER TABLE mariner ENABLE ALWAYS TRIGGER "'\''|| tgname||'\''";'\'' from pg_trigger where tgconstrname='\''personal_email_sender_marinerid_fkey'\'' and tgisconstraint and
> tgrelid='\''mariner'\''::regclass' | psql -f -
>
> and then test with:
> BEGIN ;
> set session_replication_role TO 'replica';
> DELETE FROM mariner where id = 23700;
> SELECT * from personal_email_sender where marinerid = 23700;
>  email | marinerid
> -------+-----------
> (0 rows)
>
> ^^^ and verify that the ON DELETE CASCADE is run.
>
> So, besides the obvious comment, that one should not mess with system triggers, do you see any potential gotcha with this particular one?

I think this is generally OK.

But if you are doing this, you probably have a replication system that
is running under the replica session role. The question is then whether
that replication system is OK with it. If you create replication sets
that contain parts of foreign key relationships and not others then you
might run into constraint violations.

--
Peter Eisentraut http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

In response to

Responses

Browse pgsql-admin by date

  From Date Subject
Next Message Alexandre hadjinlian guerra 2019-07-06 05:35:07 are there any expectations about pg indexes having prefix compression?
Previous Message Achilleas Mantzios 2019-07-05 10:17:05 pgsql 8.3 : force FK (and consequently ON DELETE CASCADE) to be run under session_replication_role TO 'replica'