From: Jerry Sievers <gsievers19@comcast.net>
To: Andreas Joseph Krogh <andreas@visena.com>
Cc: pgsql-performance@postgresql.org
Subject: Re: Poor performance when deleting from entity-attribute-value type
 master-table
References: <VisenaEmail.6a.6ef5c9caf9b77096.14b702ca7f3@tc7-visena>
Date: Mon, 09 Feb 2015 15:42:28 -0600
In-Reply-To: <VisenaEmail.6a.6ef5c9caf9b77096.14b702ca7f3@tc7-visena> (Andreas
 Joseph Krogh's message of "Mon, 9 Feb 2015 22:12:55 +0100 (CET)")
Message-ID: <86wq3qyd6j.fsf@jerry.enova.com>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/23.4 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Precedence: bulk
Sender: pgsql-performance-owner@postgresql.org

Andreas Joseph Krogh <andreas@visena.com> writes:

> Hi all.
> 
> Using PG-9.4.0 I'm seeing this trying to delete from an "entity"-master table:
> 
> *# explain analyze delete from onp_crm_entity where entity_id IN (select tmp.delivery_id from temp_delete_delivery_id tmp);
>  QUERY PLAN 
> ---------------------------------------------------------------------------------------------------------------------------------------------------
> Delete on onp_crm_entity (cost=0.43..5673.40 rows=1770 width=12) (actual time=7.370..7.370 rows=0 loops=1)
>  -> Nested Loop (cost=0.43..5673.40 rows=1770 width=12) (actual time=0.050..1.374 rows=108 loops=1)
>  -> Seq Scan on temp_delete_delivery_id tmp (cost=0.00..27.70 rows=1770 width=14) (actual time=0.014..0.080 rows=108 loops=1)
>  -> Index Scan using onp_crm_entity_pkey on onp_crm_entity (cost=0.43..3.18 rows=1 width=14) (actual time=0.010..0.011 rows=1 loops=108)
>  Index Cond: (entity_id = tmp.delivery_id)
> Planning time: 0.314 ms
> Trigger for constraint onp_crm_activity_entity_id_fkey: time=4.141 calls=108
> Trigger for constraint ...
> Trigger for constraint ...
> Trigger for constraint ...
> 
> 
> I have lots of tables referencing onp_crm_entity(entity_id) so I expect the poor performance of deleting from it is caused by all the triggers firing to check
> FKI-constraints.
> 
> Are there any ways around this or do people simply avoid having FKs in schemas like this?

The classic problem is that  one/more of your referring tables is
non-trivial in size and you are missing  an index on the referring column(s).

Insure that this condition does not exist before butchering your design :-)


> Thanks.
> 
> --
> Andreas Joseph Krogh
> CTO / Partner - Visena AS
> Mobile: +47 909 56 963
> andreas@visena.com
> www.visena.com
> [cid]
>

-- 
Jerry Sievers
Postgres DBA/Development Consulting
e: postgres.consulting@comcast.net
p: 312.241.7800


-- 
Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance