Re: Vacuum time degrading

On 4/4/05 8:50 AM, "Tom Lane" <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> That doesn't follow from what you said. Did you check that the physical
> sizes of the indexes were comparable before and after the reindex?

No, how do I do that (or where is it documented how to do it)?

How is it not consistent? I believe you suggested the reindex. The initial
timing was shortly after a database reload. The index would have been built
in sorted order, correct? This was the 1 hour time. After a period of
months, the index values have been inserted in random order. The DB size is
up 50% but the vacuum time is up a factor of 6+. Presumably the index is
being read by moving the heads all over the place. I reindex, and the index
is rebuilt in sorted order. Vacuum is now down to 1.5 hours - a linear
scaling from the original numbers. The locality of reference in reading the
indexes in order should be much better.

Wes