> To get a partial VACUUM ANALYZE that was actually usefully faster than
> the current code, I think you'd have to read just a few percent of the
> blocks, which means much less than a few percent of the rows ... unless
> maybe you picked selected blocks but then used all the rows in those
> blocks ... but is that a random sample? It's debatable.
>
> I find it hard to believe that VAC ANALYZE is all that much slower than
> plain VACUUM anyway; fixing the indexes is the slowest part of VACUUM in
> my experience. It would be useful to know exactly what the columns are
> in a table where VAC ANALYZE is considered unusably slow.
VACUUM ANALYZE does a huge number of adt/ function calls. It must be
those calls that make ANALYZE slower. People report ANALYZE is
certainly slower, and that is the only difference.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026