Re: database question - Mailing list pgsql-general

From Tom Lane
Subject Re: database question
Date
Msg-id 25937.1222731092@sss.pgh.pa.us
Whole thread Raw
In response to Re: database question  ("Scott Marlowe" <scott.marlowe@gmail.com>)
List pgsql-general
"Scott Marlowe" <scott.marlowe@gmail.com> writes:
> On Mon, Sep 29, 2008 at 8:21 AM,  <john.crawford@sirsidynix.com> wrote:
>> -rw------- 1 postgres root     1073741824 Sep 29 15:15 2613.77
>> -rw------- 1 postgres root     1073741824 Sep 29 15:15 2613.83
>>
>> What are these files and why have they suddenly started to be created
>> and why so large?

> PostgreSQL automatically splits table files into 1G chunks so it can
> run on OSes with file size limits.  These are part of the table
> identified by the oid 2613.  You can find it by looking in pg_class.

Actually relfilenode, not oid, is the thing to look at.  But a table
with such a small relfilenode number must be a system catalog, and a
quick look shows that in any recent PG version it's pg_largeobject.

So the answer is you've got something that's gone hog-wild on creating
large objects and not deleting them; or maybe the application *is*
deleting them but pg_largeobject isn't getting vacuumed.

            regards, tom lane

pgsql-general by date:

Previous
From: Tom Lane
Date:
Subject: Re: Can't cast from char to integer...
Next
From: Tom Lane
Date:
Subject: Re: Counting unique rows as an aggregate.