Op 01-09-11 14:43, Scott Marlowe schreef:
> On Thu, Sep 1, 2011 at 6:38 AM, Rik Bellens<rik(dot)bellens(at)telin(dot)ugent(dot)be> wrote:
>> Op 01-09-11 14:22, Scott Marlowe schreef:
>>> Yeah, could be. Take a look at this page:
>>> http://wiki.postgresql.org/wiki/Show_database_bloat and see if the
>>> query there sheds some light on your situ.
>> thanks for this answer
>>
>> if i run the query, I get 12433752064 wasted bytes on stats_count_pkey, so I
>> suppose that is the reason
> Also look into installing something like nagios and the
> check_postgresql.pl plugin to keep track of these things before they
> get out of hand.
>
> csb time: Back in the day when pg 6.5.3 and 7.0 was new and
> interesting, I had a table that was 80k or so, and an index that was
> about 100M. Back when dual core machines were servers, and 1G ram was
> an extravagance. I had a process that deleted everything from the
> table each night and replaced it, and the index was so huge that
> lookups were taking something like 10 seconds each. A simple drop /
> create index fixed it right up. The check_postgresql.pl script is a
> god sent tool to keep your db healthy and happy.
after running reindex on the stats_count_pkey, the disk size of this
index was reduced with about 2 gigs, but the size of the table itself
was still very large.
running 'vacuum full' also reduced the table size from 14 gigs to about
2 gigs.
I will now regullarly check the database with the mentioned tools and
queries.
thank you for the very usefull tips