Re: confusting results from pg_database_size

On Thursday 04 February 2010 1:52:36 am AutoVHC Dev Team wrote:
> As part of an testing an archive solution I've updated and deleted 3 or 4
> million rows in different tables. I wanted to see how much this shrunk the
> database size by running, so I ran a VACUUM FULL FREEZE ANALYZE on both
> databases - this took a long time, which is ok considering the number of
> rows I updated/deleted
>
> To get the db size I ran the following:
>
> SELECT pg_size_pretty(pg_database_size('deleted_rows_db')),
> pg_size_pretty(pg_database_size('original_db'));
>
> Obviously you'd expect the database that is minus 3 million rows to be
> smaller, however it was 300Mb larger. (This is a test environment where no
> one else has access to the database, both databases were restored from the
> same backup. Running a count on the tables shows the rows have definitely
> been deleted. - This is the second time I've done this as I considered that
> I must have done something wrong the first time).
>
> Does anyone know why the database with fewer rows is larger? how I can find
> the real size?, or do something to the database to lose this bloat? (It is
> as if the vacuum didn't work - though it appeared to).
>
> Thanks in advance.
>
> -Dan Shoubridge

Do you have indexes on the tables? If so see here:

http://www.postgresql.org/docs/8.4/interactive/routine-vacuuming.html
"Another disadvantage of VACUUM FULL is that while it reduces table size, it
does not reduce index size proportionally; in fact it can make indexes larger.
Generally, therefore, administrators should strive to use standard VACUUM and
avoid VACUUM FULL. '

--
Adrian Klaver
adrian(dot)klaver(at)gmail(dot)com