Re: running only pg_autovacuum for one week

Scott Marlowe wrote:
> On Mon, 2004-08-02 at 07:08, Gaetano Mendola wrote:
>
>>-----BEGIN PGP SIGNED MESSAGE-----
>>Hash: SHA1
>>
>>Hi all,
>>I did the experiment of running only the autovacuum for one week without
>>running my daily autovacuum full and reindex on heavy updated/inserted
>>tables.
>>Yesterday I reenabled the vacuum full and reindex and, as you can see
>>from the attachment, I recover 600MB of wasted space.
>
>
> autovacuum is a daemon. You don't run it once a week, you set it loose
> and forget about it. Unless you're running it then shutting it down,
> running once a week is unnecessary.

Sorry, I meant that I had running "only" the pg_autovacuum for the entire week,
without run also my vacuum full + reindex once a day. I was only stopping it and
rerunning for logrotation purpose.
Analyzing the graph on Wednedsay I had a few of processes in "idle in transaction"
state and as you can see the graph had a big ramp and that space was not reclamed
till this morning :-(

> Also, what are you fsm settings in the postgresql.conf file?

At the end of my autovacuum full I have:

INFO: free space map: 603 relations, 38202 pages stored; 40592 total pages needed
DETAIL: Allocated FSM size: 1000 relations + 2000000 pages = 11780 kB shared memory.

So I think that I'm not reaching the limits.

As explained in another thread untill I can not set the "threasholds" per table
the autovacuum is useless ( milions rows tables with hundred of insert per day ).

Regards
Gaetano Mendola