On May 23, 2007, at 9:26 AM, Susan Russo wrote:
> I've played 'catch up' wrt adjusting max_fsm_pages (seems to be a
> regular event),
> however am wondering if the vacuum analyze which reports the error was
> actually completed?
Yes, it completed. However not all pages with open space in them are
accounted for, so you will probably end up allocating more pages than
you otherwise would have. I take it as a sign of not running vacuum
often enough so that the pages with available space get filled up
sooner.
I'd bump fsm pages up to perhaps double or triple what you've got
now, and try running vacuum a bit more often on your hottest tables
(or even consider changing your table structure to limit the
"hotness" of that table). For example, if the table has a lot of
columns, but only one is updated often, split that column out into
its own table so that you have all the changes clustered into the
same set of pages while leaving the rest of the columns in place.
