On Mon, Jan 16, 2012 at 5:09 AM, Noah Misch <noah@leadboat.com> wrote:
>
> pgstattuple()'s decision to treat half-dead pages like deleted pages is
> better. That transient state can only end in the page's deletion.
>
the only page in that index has 200 records (all live 0 dead) using
half the page size (which is a leaf page and is not half dead, btw).
so, how do you justify that pgstattuple say we have just 25% of free
space?
postgres=# SELECT * from bt_page_stats('pgbench_tellers_pkey', 1);
-[ RECORD 1 ]-+-----
blkno | 1
type | l
live_items | 200
dead_items | 0
avg_item_size | 16
page_size | 8192
free_size | 4148
btpo_prev | 0
btpo_next | 0
btpo | 0
btpo_flags | 3
> I don't know about counting non-leaf pages
ignoring all non-leaf pages still gives a considerable difference
between pgstattuple and relation_free_space()
--
Jaime Casanova www.2ndQuadrant.com
Professional PostgreSQL: Soporte 24x7 y capacitación