On Mon, Nov 23, 2009 at 5:25 PM, Christopher Browne
<cbbrowne at ca.afilias.info> wrote:
>
> Well, Slony-I does invalidate quite a lot of pg_listener tuples, so it's
> not a surprise for this to happen if pg_listener isn't getting vacuumed.
>
> The cleanup thread is supposed to vacuum pg_listener once in a while
> (I'd expect about every 10 minutes, if slon runs with defaults), so I'd
> wonder why that, or some other vacuum, is not happening.

Nope, i think there is a problem here.
here is a vacuum verbose analyze pg_listener on this specific slave
that lagged :

INFO:  vacuuming "pg_catalog.pg_listener"
INFO:  "pg_listener": removed 160 row versions in 7 pages
INFO:  "pg_listener": found 160 removable, 4 nonremovable row versions
in 20129 pages
DETAIL:  0 dead row versions cannot be removed yet.
There were 1609042 unused item pointers.
20129 pages contain useful free space.
0 pages are entirely empty.
CPU 0.00s/0.04u sec elapsed 0.04 sec.
INFO:  analyzing "pg_catalog.pg_listener"
INFO:  "pg_listener": scanned 20129 of 20129 pages, containing 4 live
rows and 0 dead rows; 4 rows in sample, 4 estimated total rows
VACUUM

1.6 millions unused item pointers.
20kpage * 8KB -> 160MB !!!!wtf!omg!bbq!!!!111oneone1!1oneoneeleven

I will check what you said.
Thx.

-- 
F4FQM
Kerunix Flan
Laurent Laborde