On Wed, Sep 25, 2002 at 06:46:27PM -0400, Tom Lane wrote:
...
> That's about what I'd expect it to be doing. The final decision to exit
> would normally be made when we see the shutdown process exit (about
> line 1587 in postmaster.c). What are the contents of ShutdownPID,
> CheckPointPID, Shutdown, and FatalError?
(gdb) print ShutdownPID
$1 = 0
(gdb) print CheckPointPID
$2 = 0
(gdb) print Shutdown
$3 = 2
(gdb) print FatalError
$4 = 0 '\000'
> Are there any remaining child
> processes of the postmaster?
# ps ax | grep post
10828 p3 S+ 0:00.04 postmaster -o -W 15 -d4 (postgres)
10829 p3 S+ 0:00.00 postmaster: stats buffer process (postgres)
10831 p3 S+ 0:00.01 postmaster: stats collector process (postgres)
11387 p5 D+ 0:00.00 grep post
11360 p9 SW+ 0:00.00 vi postmaster.c
> That tip is pretty obsolete, but before you pull the trigger it would be
> nice to try to learn more. I wonder if you have hit some obscure race
> condition that prevents the postmaster from realizing it's done?
? Nothing else was going on... Now I'm hammering the box looking for the
constraint in the 3.3Gb dump file, so swap is a bit low.
Cheers,
Patrick