Tom Lane wrote:
> Maxim Boguk <mboguk@masterhost.ru> writes:
>> [ ndistinct estimates way off ]
>
> Estimating the number of distinct values in a distribution with a long
> tail is just a really hard problem :-(
If I have a table where I know it has this property, is there
any way I can tell autovacuum's ANALYZE to go ahead and scan
the whole table? ISTM a vacuum_cost_delay throttled full
table scan might be less harmful then a bad ndistinct guess.