On Wed, Dec 9, 2009 at 5:16 AM, David Richardson
<..hidden..> wrote:
Just to stick in my $0.02: I have had this type of semi-random
"collision" problem (MS-SQL, however) due to hashed indices when the
hashed values collide. Not sure if it would help here, but you might
be a little careful about replacing stuff if they are using this sort
of index - a little double-check in SQL code might be in order..
In PG, this is USUALLY caused by a sequence somehow getting out of whack. The keys are indexed though so I suppose that could be a factor as well.
If the setval statement doesn't fix it (by setting the sequence to the maximum value in the table), then a reindex might be in order.
Best Wishes,
Chris Travers