Significantly larger toast tables on 8.4? - Mailing list pgsql-hackers

From Alex Hunsaker
Subject Significantly larger toast tables on 8.4?
Date
Msg-id 34d269d40901012030x422a98fan40d12746a6c3088f@mail.gmail.com
Whole thread Raw
Responses Re: Significantly larger toast tables on 8.4?  ("Alex Hunsaker" <badalex@gmail.com>)
Re: Significantly larger toast tables on 8.4?  ("Alex Hunsaker" <badalex@gmail.com>)
List pgsql-hackers
I just did a fresh import of my alpha database into 8.4 and noticed
that the size of the database had increased significantly:

8.4: 6.3G
8.3: 4.9G

Tracking it down the main difference seems to a toast tables namely this one:

ls -ltrh 8.3/base/16516/430156
-rw------- 1 postgres postgres 145M 2009-01-01 00:12 8.3/base/16516/430156

and the same toast table in 8.4:

ls -ltrh 8.4/base/16513/347706
-rw------- 1 postgres postgres 967M 2009-01-01 20:56 8.4/base/16513/347706

This table consists mainly of perl Storable binary blobs in a bytea
column schema looks like:
   Column     |           Type           | Modifiers
---------------+--------------------------+---------------------------------------------------------------------action
     | text                     |die_id | integer                  | not null default
 
nextval('dies_die_id_seq'::regclass)cparam        | bytea                    |date_created  | timestamp with time zone
|not null default now()db_profile    | bytea                    |debug         | bytea                    |defunct
| smallint                 | not null default 0env           | bytea                    |login         | bytea
         |msg           | text                     |open_user_id  | integer                  |page_load_id  | integer
              |session_id    | integer                  |state         | bytea                    |state_action  |
bytea                   |user_id       | integer                  |whoops        | bytea                    |
 
Indexes:  "dies_pkey" PRIMARY KEY, btree (die_id)


My hunch is its related to
http://git.postgresql.org/?p=postgresql.git;a=commit;h=3b6942b3f2fe733572c05a71cb2d12e5ece60cdbits
or for the CVS inclined
http://archives.postgresql.org/pgsql-committers/2008-03/msg00121.php

But if anything that looks like it should help reduce size...

Ideas?


pgsql-hackers by date:

Previous
From: "Robert Haas"
Date:
Subject: Re: posix_fadvise v22
Next
From: "Alex Hunsaker"
Date:
Subject: Re: Significantly larger toast tables on 8.4?