Re: Postgresql 'eats' all mi data partition

Tomas Szepe <szepe(at)pinerecords(dot)com> writes:
> INFO: --Relation public.stats_min--
> INFO: Index stats_min_start: Pages 34445; Tuples 1985464: Deleted 404651.
> CPU 1.19s/2.41u sec elapsed 17.86 sec.
> INFO: Index stats_min_pkey: Pages 76501; Tuples 1986938: Deleted 404651.
> CPU 3.98s/5.47u sec elapsed 217.07 sec.
> INFO: Removed 404651 tuples in 6118 pages.
> CPU 0.83s/0.77u sec elapsed 13.52 sec.
> INFO: Pages 25295: Changed 4615, Empty 0; Tup 1985464: Vac 404651, Keep 0, UnUsed 468220.
> Total CPU 7.19s/8.78u sec elapsed 252.67 sec.
> INFO: --Relation public.stats_hr--
> INFO: Index stats_hr_start: Pages 57654; Tuples 10811294: Deleted 348991.
> CPU 3.09s/5.27u sec elapsed 63.67 sec.
> INFO: Index stats_hr_pkey: Pages 78301; Tuples 10814527: Deleted 348991.
> CPU 5.11s/6.39u sec elapsed 152.78 sec.
> INFO: Removed 348991 tuples in 8333 pages.
> CPU 1.09s/0.92u sec elapsed 36.46 sec.
> INFO: Pages 217213: Changed 1362, Empty 0; Tup 10810476: Vac 348991, Keep 0, UnUsed 352822.
> Total CPU 17.09s/13.98u sec elapsed 284.52 sec.

You've definitely got an index-bloat problem on stats_min (the indexes
are four times the size of the table :-(), and I suspect the same on
stats_hr, though not quite as bad. What are the datatypes of the
index columns?

> Hmm, you seem to suggest that we might expect a change in this regard
> as 7.4 ships. Is that right?

7.4 should improve matters.

regards, tom lane