On Wed, 2005-11-09 at 14:20 -0500, Tom Lane wrote:
> Rod Taylor <pg@rbt.ca> writes:
> > As you can see, we have duplicates within the table (heap) of a primary
> > key value. The index itself only references one of these tuples.
>
> Can you put together a test case to reproduce this? It doesn't have to
> fail every time, as long as it fails once in awhile ...
Seems not. I've done millions of iterations of the same type of
functionality that happens with these structures and haven't produced a
single case. These are fairly low usage structures, so I think I've done
about 3 months worth of work, which in production had 20 bad tuples. I
tried playing with various delays, vacuum schedules, and number of
parallel processes.
Whatever is happening is from interaction not contained within the
structures showing the symptoms.
I'll watch it a bit closer now that I know the problem exists to see if
I can find a pattern.
--