On Tue, 2007-08-07 at 16:52 -0400, Tom Lane wrote:
> "Simon Riggs" <simon@2ndquadrant.com> writes:
> > On Tue, 2007-08-07 at 15:14 -0400, Tom Lane wrote:
> >> "Simon Riggs" <simon@2ndquadrant.com> writes:
> >>> We also need something that will re-zero the stats when they reach
> >>> anywhere near integer overflow, since we must not allow them to wrap. I
> >>> would suggest we simply reset all values to zero for that table.
> >>
> >> pgstat counters are int64.
>
> > You would prefer undefined behaviour at wrap?
>
> You should live so long as to have a problem with it. Do the math:
> at one increment every nanosecond, 24x7x365, you'd be risking overflow
> after about 300 years of continuous initdb-less operation. For someone
> opining that important features are OK to omit from HOT for 8.3, I have
> to question your judgment in worrying about this.
I'm not worried about it, but I was mistaken in thinking you might be.
If you're OK with HOT as-is, then sure, I'll have partial indexes too
and much more besides. The question is: are you OK with HOT as-is?
Should we take it further? How far?
-- Simon Riggs EnterpriseDB http://www.enterprisedb.com