Re: possible vacuum improvement?

On 3 Sep 2002 at 9:36, Mario Weilguni wrote:
> That is not really practicable, one datebase has 107 tables, and making a
> cron job
> with 107 vacuum calls is completly out of question and very error prone
> anyway.

That's correct.. What are the possible alternatives? Either backend has to
support something or the DBA has to script something.

1)If number of tables that need vacuum are far more than those who don't, then
a simple all vacuum would do. But again sizes of individual tables will affect
that judgement as well.

2)As OP suggested, if vacuum could pick up only those tables marked by
bitfields, ay by an additional option like, 'vacuum analyse frequent_ones'..
this is going to need a backend change.

3)I guess scripting cron job for vacuum is one time job. If it's desparately
needed, say 60 tables out of 107 require vacuum, personally I would spend some
time making that script. Depends upon the requirement actually.

On a sidenote, does anybody have some statistics from benchmark may be, as in
what's a rule of thumb for vacuuming? I found that a vacuum every 5K-10K
transactions increases the tps like anything but below 1K transactions, it's
not as much effective. May be one should consider this factor as well..

Bye
Shridhar

--
Pascal: A programming language named after a man who would turn over in his
grave if he knew about it. -- Datamation, January 15, 1984