Bruce Momjian <pgman@candle.pha.pa.us> writes:
> The one problem with the signal approach is how long does the system
> wait before giving up on the app shutdown? Seems that should be
> something controllable by the admin, but without shutdown scripts, it
> isn't.
I believe 20 seconds is the standard number --- that's plenty for
Postgres. (I know that it is about 20 seconds on OS X, because
that's how much time tended to get added to the shutdown procedure
back when OS X 10.0 had that shutdown bug that prevented the postmaster
from forking a shutdown subprocess.)
The fact that the number isn't readily configurable is indeed a PITA.
In a previous lifetime I ran a data-collection application that needed
more than 20 seconds to shut down, and so would not exit cleanly if
you didn't have a shutdown script step that would wait for it. But
I don't see it as a problem for Postgres.
regards, tom lane