Alvaro Herrera wrote:
> You missed putting back the BUG comment that used to be there about
> this.
>
This was deliberate, I did mention the condition in the comment at
the beginning of the file. This actually makes it a feature :)
Seriously though, do you think that this is still a problem? Given
the rare occurrence of the revacuum and the fact that it is made
cheap by visibility map? In my initial testing, I couldn't reproduce
the revacuum. But I'll keep at it.
> In other words I think this is a bad idea, because there is a very wide
> window for a table to be vacuumed twice. Since naptime can be
> arbitrarily large, this is an arbitrarily large bug. I'm sure there are
> other ways to fix this, but please propose those before this patch.
>
I was wondering that maybe the stats subsystem shouldn't be used for
vacuum tracking at all. It maybe convenient to use, but has several
deficiencies (pobig file, lossy, no crash safety, etc). Could we move
vacuum tracking to pg_class instead?
regards,
Martin