Re: Intermittent regression test failure from index-only scans patch

Magnus Hagander <magnus(at)hagander(dot)net> writes:
> On Sun, Oct 9, 2011 at 06:34, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> Robert Haas <robertmhaas(at)gmail(dot)com> writes:
>>> Should we have another counter for heap fetches avoided? Seems like that could be useful to know.

>> Hm. I'm hesitant to add another per-table (or per index?) statistics
>> counter because of the resultant bloat in the stats file.

> We certainly need *some* way to figure out if this has been used,
> IMHO. So yeah, if the current way doesn't scale enough, we need to
> think of some other way. But I'm not sure one more counter would
> really bloat it that bad? OTOH, repeating that reasoning enough time
> will eventually make it enough to care about...

You can already tell whether it's happening by comparing idx_tup_read
versus idx_tup_fetch. Now that measure does conflate some things, like
whether the tuple was not read at all or was read and rejected as not
visible, but I'm not at all convinced that another counter is worth its
weight. If invisible tuples are a significant part of the table then
index-only scanning isn't going to be very useful to you anyway.

regards, tom lane