pg.ddx.io  pgsql-performance@postgresql.org mailing list archive  
help / color / mirror / Atom feed
From: Junaili Lie <junaili@gmail.com>
To: pgsql-performance@postgresql.org
Subject: slow deletes on pgsql 7.4
Date: Tue, 25 Apr 2006 14:41:03 -0700
Message-ID: <8d04ce990604251441ya85c42evb47946272536cabe@mail.gmail.com> (raw)

Hi all,
we encounter issues when deleting from a table based on id (primary key). On
certain 'id', it took forever to delete and the i/o is 100% busy.
Table scenario has around 1400 entries. It is the parent of 3 other table.
                                 Table "public.scenario"
     Column      |         Type          |                   Modifiers
-----------------+-----------------------+------------------------------------------------
 id              | bigint                | not null default
nextval('scenario_seq'::text)
 name            | character varying(50) |
 description     | text                  |
 subscriber_id   | bigint                |
 organization_id | bigint                |
 schedule_id     | bigint                |
Indexes:
    "scenario_pkey" primary key, btree (id)
    "org_ind_scenario_index" btree (organization_id)
    "sch_ind_scenario_index" btree (schedule_id)
    "sub_ind_scenario_index" btree (subscriber_id)
Check constraints:
    "$3" CHECK (schedule_id >= 0)
    "$2" CHECK (organization_id >= 0)
    "$1" CHECK (subscriber_id >= 0)
Foreign-key constraints:
    "0_4774" FOREIGN KEY (schedule_id) REFERENCES schedule(id) ON DELETE
CASCADE
    "0_4773" FOREIGN KEY (organization_id) REFERENCES organization(id) ON
DELETE CASCADE
    "0_4772" FOREIGN KEY (subscriber_id) REFERENCES subscriber(id) ON DELETE
CASCADE

In all the child tables, the foreign key has the same data type and are
indexed.
When I do "delete from scenario where id='1023', it takes less than 200 ms.
But when i do "delete from scenario where id='1099', it took forever (more
than 10 minutes that i decided to cancel it.
I can't do explain analyze, but here is the explain:
MONSOON=# begin;
BEGIN
MONSOON=# explain delete from scenario where id='1099';
                                  QUERY PLAN
------------------------------------------------------------------------------
 Index Scan using scenario_pkey on scenario  (cost=0.00..3.14 rows=1
width=6)
   Index Cond: (id = 1099::bigint)
(2 rows)

MONSOON=# explain delete from scenario where id='1023';
                                  QUERY PLAN
------------------------------------------------------------------------------
 Index Scan using scenario_pkey on scenario  (cost=0.00..3.14 rows=1
width=6)
   Index Cond: (id = 1023::bigint)
(2 rows)

MONSOON=# explain analyze delete from scenario where id='1023';
                                                       QUERY PLAN
------------------------------------------------------------------------------------------------------------------------
 Index Scan using scenario_pkey on scenario  (cost=0.00..3.14 rows=1
width=6) (actual time=0.028..0.030 rows=1 loops=1)
   Index Cond: (id = 1023::bigint)
 Total runtime: 0.174 ms
(3 rows)

I have also tried increasing statistics on both parent and child tables to
100, vacuum analyze parent and all child tables. But still the same
slowness.
The o/s is Solaris 10, with fsync = true.
Any ideas what's going on?
Thanks in advance,

J

view thread (6+ messages)  latest in thread

Message-ID: <8d04ce990604251441ya85c42evb47946272536cabe@mail.gmail.com>
Permalink:  ../8d04ce990604251441ya85c42evb47946272536cabe@mail.gmail.com/
Also on:    postgresql.org/message-id/8d04ce990604251441ya85c42evb47946272536cabe@mail.gmail.com

 · 

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: pgsql-performance@postgresql.org
  Cc: junaili@gmail.com
  Subject: Re: slow deletes on pgsql 7.4
  In-Reply-To: <8d04ce990604251441ya85c42evb47946272536cabe@mail.gmail.com>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox