On Mon, Oct 13, 2008 at 6:44 AM, Gregory Stark <stark@enterprisedb.com> wrote:
> How much memory the OS allows Postgres to allocate will depend on a lot of
> external factors. At a guess you had some other services or queries running at
> the same time the first time which reduced the available memory.
I'm sorry- I was insufficiently clear. Postgres was the only service
running, and there were no additional queries happening at the same
time. (This database is on a dedicated machine; the only other things
that run are some decision-support applications that were all off at
the time.) In addition, the 35% memory usage number was for user-space
processes in total, not for postgres specifically; the swap space was
completely clear. maintenance_work_mem + work_mem is well under the
total amount of RAM on the system, and certainly well under RAM +
swap.
I'll give a try to building that index with a lower
maintenance_work_mem this evening when I can shut off the other
processes again, though given the above it strikes me as unlikely to
be the problem.
Also, the thing that has me even more confused is the fact that it
worked when I added an additional column to the index.
--
- David T. Wilson
david.t.wilson@gmail.com
