"Benjamin Krajmalnik" <kraj@illumen.com> writes:
> I was playing around trying to tweak the performance of the database.
> As part of the process, I increased the number of WAL buffers to 32 and
> the checkpoint segments to 16.
> That appears to have been the culprit! I set them back at the same
> levels I was previously using (16 and 4, respectively), and now the
> table is no longer misbehaving.
Um ... changing those parameters should have affected the contents of
$PGDATA/pg_xlog, but not of the table per se. What were you measuring
exactly?
regards, tom lane