Re: index bloat

> -----Original Message-----
> From: Tom Lane [mailto:tgl(at)sss(dot)pgh(dot)pa(dot)us]
> Sent: Tuesday, July 12, 2005 12:26 PM
>
> "David Esposito" <pgsql-general(at)esposito(dot)newnetco(dot)com> writes:
> > As promised, here are two runs of VACUUM VERBOSE on the
> problem table ...
>
> BTW, the tail of the VACUUM VERBOSE output ought to have
> something about
> overall usage of the FSM --- what does that look like?
>

When I was first looking into this problem, I googled around a bit and
saw that there is a message that's printed when the fsm_pages or
fsm_relations is too low ... but I just double checked and can't find
the phrase "max_fsm_pages" in any of my logs from the past 2 months (i
only run postgres with log_min_error_statement of NOTICE so if it gets
logged as an INFO, i probably wouldn't have it in the log files) ...

Saturday 1:30am

INFO: free space map: 525 relations, 125128 pages stored; 136480 total
pages needed
DETAIL: Allocated FSM size: 10000 relations + 1000000 pages = 6511 kB
shared memory.

Monday 1:30am

INFO: free space map: 525 relations, 168816 pages stored; 171904 total
pages needed
DETAIL: Allocated FSM size: 10000 relations + 1000000 pages = 6511 kB
shared memory.

Tuesday 1:30am

INFO: free space map: 528 relations, 172357 pages stored; 170096 total
pages needed
DETAIL: Allocated FSM size: 10000 relations + 1000000 pages = 6511 kB
shared memory.