Re: Query performance on session table

Here is the results of VACUUM VERBOSE and a query executed right after
vacuum is complete. Still 30 sec for 700 rows. Do I need to do VACUUM
FULL?

Thanks,

Burak

# VACUUM VERBOSE session;
INFO: vacuuming "public.session"
INFO: index "session_pkey" now contains 854 row versions in 5255 pages
DETAIL: 9212 index row versions were removed.
5028 index pages have been deleted, 5004 are currently reusable.
CPU 0.09s/0.03u sec elapsed 4.65 sec.
INFO: "session": removed 9212 row versions in 3086 pages
DETAIL: CPU 0.08s/0.15u sec elapsed 1.47 sec.
INFO: "session": found 9212 removable, 793 nonremovable row versions
in 373746 pages
DETAIL: 34 dead row versions cannot be removed yet.
There were 2938041 unused item pointers.
0 pages are entirely empty.
CPU 2.28s/0.82u sec elapsed 68.88 sec.
INFO: "session": truncated 373746 to 147910 pages
DETAIL: CPU 2.65s/0.51u sec elapsed 37.76 sec.
INFO: vacuuming "pg_toast.pg_toast_123221198"
INFO: index "pg_toast_123221198_index" now contains 0 row versions in 1 pages
DETAIL: 0 index pages have been deleted, 0 are currently reusable.
CPU 0.00s/0.00u sec elapsed 0.02 sec.
INFO: "pg_toast_123221198": found 0 removable, 0 nonremovable row
versions in 0 pages
DETAIL: 0 dead row versions cannot be removed yet.
There were 0 unused item pointers.
0 pages are entirely empty.
CPU 0.00s/0.00u sec elapsed 0.02 sec.
VACUUM

citizenre=# EXPLAIN ANALYZE SELECT * FROM session;
QUERY PLAN
-----------------------------------------------------------------------------------------------------------------------
Seq Scan on session (cost=0.00..147917.93 rows=793 width=282)
(actual time=27260.391..29186.760 rows=708 loops=1)
Total runtime: 29187.556 ms
(2 rows)

On 6/28/07, Michael Glaesemann <grzm(at)seespotcode(dot)net> wrote:
>
> On Jun 28, 2007, at 15:26 , Burak Seydioglu wrote:
>
> > I am having performance issues with a table that holds session_data.
> > This table is heavily updated and daily vacuumed.
>
> If it's heavily updated, vacuuming once a day may not be enough. You
> can elect to vacuum a single table.
>
> What's the output of VACUUM VERBOSE session ?It should tell you how
> many dead tuples were removed.
>
> You may also want to consider using autovacuum to help keep your
> session table fairly clean.
>
> Michael Glaesemann
> grzm seespotcode net
>
>
>