Laurent Laborde wrote:
> Sooo ...
>
> i disabled autovacuum  using the following request :
> INSERT INTO pg_catalog.pg_autovacuum (vacrelid, enabled,
> vac_base_thresh, vac_scale_factor, anl_base_thresh, anl_scale_factor,
> vac_cost_delay, vac_cost_limit, freeze_min_age, freeze_max_age) SELECT
> oid, 'f', -1, -1, -1, -1, -1, -1, -1, -1 FROM pg_catalog.pg_class
> WHERE relnamespace = (SELECT OID FROM pg_namespace WHERE nspname =
> '_ob2replication' ) AND relhasindex;
>
> I waited ~24h then ran a vacuum verbose analyze on some slony table.
>
> INFO:  "sl_listen": found 0 removable, 1282 nonremovable row versions
> in 30 pages
> DETAIL:  0 dead row versions cannot be removed yet.
> There were 6113 unused item pointers.
> 27 pages contain useful free space.
> 0 pages are entirely empty.
>
> INFO:  "sl_event": found 0 removable, 2599 nonremovable row versions
> in 7229 pages
> DETAIL:  0 dead row versions cannot be removed yet.
> There were 784524 unused item pointers.
> 7220 pages contain useful free space.
> 0 pages are entirely empty.
>
> INFO:  "sl_confirm": found 0 removable, 27585 nonremovable row
> versions in 321 pages
> DETAIL:  0 dead row versions cannot be removed yet.
> There were 65301 unused item pointers.
> 188 pages contain useful free space.
> 0 pages are entirely empty.
>
> INFO:  "sl_log_1": found 32843 removable, 16694 nonremovable row
> versions in 4464 pages
> DETAIL:  0 dead row versions cannot be removed yet.
> There were 151138 unused item pointers.
> 3648 pages contain useful free space.
> 0 pages are entirely empty.
>
>
> Look so bad ?
> However... i can't find anything in the log saying that slony is
> manually vacuuming.
> Should i restart slony ?
>
>   
Nothing looks notably wrong, here.

There's a fair bit of dead data in sl_log_1, but not *enormous*.  The 
log rotation approach used in the modern versions of Slony-I (1.2 and 
newer) means that even if you never vacuum sl_log_1 or sl_log_2, that 
data will eventually get truncated away.

Actually, in 2.0, all we ever do with sl_log_{1,2} is to truncate them.  
They should be pretty useless to vacuum!

You might want to run CLUSTER or VACUUM FULL+reindex on several of the 
other tables, as they have bloated to fairly considerable size, and an 
ordinary vacuum won't quickly do anything about that.  Once they've 
bloated, you need to do a bit of maintenance to fix that :-(.