Re: RFE: Make statistics robust for unplanned events

Peter Geoghegan <pg(at)bowt(dot)ie> writes:
> We already *almost* pay the full cost of durably storing the
> information used by autovacuum.c's relation_needs_vacanalyze() to
> determine if a VACUUM is required -- we're only missing
> new_dead_tuples/tabentry->n_dead_tuples. Why not go one tiny baby step
> further to fix this issue?

Definitely worth thinking about, but I'm a little confused about how
you see this working. Those pg_class fields are updated by vacuum
(or analyze) itself. How could they usefully serve as input to
autovacuum's decisions?

> Admittedly, storing new_dead_tuples durably is not sufficient to allow
> ANALYZE to be launched on schedule when there is a hard crash. It is
> also insufficient to make sure that insert-driven autovacuums get
> launched on schedule.

I'm not that worried about the former case, but the latter seems
like kind of a problem.

regards, tom lane