On Sun, Aug 26, 2012 at 5:46 AM, Liron Shiri <lirons@checkpoint.com> wrote:
> Hi,
>
>
>
> We have a table which its TOAST table size is 66 GB, and we believe should
> be smaller.
>
> The table size is 472 kb. And the table has 4 columns that only one of them
> should be toasted.
>
>
>
> The table has only 8 dead tuples, so apparently this is not the problem.
>
>
>
> This table contains a column with bytea type data (kept as TOAST). We tried
> to check what is the size of the toasted data in each row by using the
> following query (the data_blob is the bytea column):
>
>
>
> SELECT nid, octet_length(data_blob) FROM my_table ORDER BY
> octet_length(data_blob) DESC;
>
>
>
> This result contain 1782 rows. The sizes I get from each row are between
> 35428 to 42084.
>
>
>
> 1782 * 38000 = 67716000 byte = 64.579 MB .
>
>
>
> What can be the reason for a table size of 66 GB? What else should I check?
Is the size of the database continuing to grow over time, or is it stable?
Have you done a hot-standby promotion on this database, perchance? I
have an open bug report on an unusual situation that began after that:
http://archives.postgresql.org/pgsql-bugs/2012-08/msg00108.php
--
fdr
