Ron Mayer <rm_pg(at)cheapcomplexdevices(dot)com> writes:
> On Mon, 28 Feb 2005, Tom Lane wrote:
>> You could shorten the intervals for which the lock is held by reducing
>> vacuum_mem, but this might be counterproductive overall.
> Does this work?
[ thinks about it... ] No, probably not; sorry for the misinformation.
Cutting vacuum_mem will reduce the number of index tuples that are to be
deleted during any one scan of the index. But if the index is large,
it's probably the scanning time and not the deletion time that is dominant.
regards, tom lane