On 2012-02-22 16:29, Tom Lane wrote:
> VACUUM ANALYZE
> consists of two separate passes, VACUUM and then ANALYZE, and the second
> pass is going to be "random" I/O by your definition no matter what.
I don't suppose there's a case where the VACUUM (1) gets to delete lots
and lots of rows that then don't need ANALYZE'ing, and (2) can do so
without actually touching all those pages?
In response to
pgsql-hackers by date
|Next:||From: Andrew Dunstan||Date: 2012-02-22 18:00:55|
|Subject: determining a type oid from the name|
|Previous:||From: Greg Smith||Date: 2012-02-22 17:36:28|
|Subject: Re: pg_test_timing tool for EXPLAIN ANALYZE overhead|