On Thu, Jul 24, 2008 at 9:06 AM, Simon Riggs <simon@2ndquadrant.com> wrote:
> I suspect this is not the root problem, but one solution to it.
Agreed. It is not the root problem. However, until DSM is fully
implemented and working, not having the ability to gather statistics
during long vacuums is problematic. Of course, you can try and
optimize vacuum by minimizing relation sizes using partitioning, but
that doesn't work in all cases.
> If the stats need such frequent updating, then the code that handles the
> stats probably needs extension/improvement to avoid such sensitivities.
Agreed, the optimizer's sensitivity to statistics in certain query
conditions is the root problem.
--
Jonah H. Harris, Senior DBA
myYearbook.com