On Mon, 2009-10-19 at 15:09 -0400, Brad Nicholson wrote:
> On Mon, 2009-10-19 at 15:01 -0400, Tom Lane wrote:
> > Brad Nicholson <bnichols@ca.afilias.info> writes:
> > > autoanalyze will automatically analyze new tables when they don't have
> > > stats. It seems logical that it should handle this case where the table
> > > also does not have stats.
> >
> > It will autoanalyze once a sufficient number of inserts have occurred.
> > The effect of a crash is just to reset the inserts-since-last-analyze
> > counters. You really haven't made the case why that's so awful that
> > we need to do things that are unpleasant along other dimensions in order
> > to avoid it. (The only ways of avoiding it that I can see would result
> > in very significant overhead added to the stats collection mechanism.)
>
> What about the question that Scott and I both asked - what about query
> plans. Will they be affected?
>
> If so, you could have a table that does not get updated, that no longer
> generates relevant query plans, and won't until manual intervention.
>
Actually, isn't there also a condition (although unlikely to occur) here
that could result in transaction wraparound?
--
Brad Nicholson 416-673-4106
Database Administrator, Afilias Canada Corp.