Peter Geoghegan wrote:
> On Thu, Feb 18, 2016 at 4:53 PM, Tom Lane wrote:
> >> there are usage patterns where half-dead pages might accumulate.
> >
> > Other than a usage pattern of "randomly SIGKILL backends every few
> > seconds", I don't see how that would happen.
>
> I meant where pages coul
On Fri, Feb 19, 2016 at 12:05 PM, Alvaro Herrera
wrote:
> But those pages are supposed to be used as the index grows. So unless
> they are forgotten by the FSM, they shouldn't accumulate. (Except where
> the table doesn't grow but only shrinks, so there's no need for new
> index pages, but I don
On Thu, Feb 18, 2016 at 4:53 PM, Tom Lane wrote:
> Only a physical-order scan, ie vacuum, would visit a dead page
> (ignoring transient corner cases like a page getting deleted while an
> indexscan is in flight to it). So I think treating it as part of the
> fragmentation measure is completely wr
