Joe Conway <mail@joeconway.com> writes:
>> Shouldn't the reference to work_mem here:
>>
>> "Allow in-memory sorts to use their full memory allocation (Jeff
>> Janes)"
>> http://www.postgresql.org/docs/9.3/interactive/release-9-3.html#AEN114956
>>
>> actually be referencing maintenance_work_mem?
> Or I suppose, rather, shouldn't it be referring to both?
In principle I suppose someone might've set maintenance_work_mem with an
eye on the space needed to build specific indexes, but it seems relatively
unlikely compared to the work_mem scenario. I think it'd just confuse
people to mention both parameters.
regards, tom lane