Comment # 29 on bug 906900 from Thomas Blume

Did some intensive debugging today.
Finally, I've found that the stop command in the init script was not even
executed because:

if pg_ctl status -s -D $DATADIR >/dev/null

fails with:

pg_ctl: no server running

Still, the shutdown log shows that there were postgres processes killed:

[  308.499821] systemd[1]: Child 1407 (postgres) died (code=killed,
status=9/KILL)
[  308.513468] systemd[1]: Child 1412 (postgres) died (code=killed,
status=9/KILL)
[  309.590384] systemd[1]: Received SIGCHLD from PID 1433 (postgres).
[  309.590467] systemd[1]: Child 1433 (postgres) died (code=killed,
status=9/KILL)
[  309.824166] systemd[1]: Received SIGCHLD from PID 1429 (postgres).
[  309.824209] systemd[1]: Child 1429 (postgres) died (code=killed,
status=9/KILL)

Michael, I need to check wheter it is the same on your machine.
Can you please add the command:

 pg_ctl status -s -D $DATADIR > /tmp/pg_ctl-status.out

to /etc/init.d/postgresql 

in the stop section below:

-->--
    stop)
        echo -n "Shutting down PostgreSQL $VERSION "
--<--

In addition, please modify the 'rc_status -v' command at the end of the stop
code to:

rc_status -v &> /tmp/rc_status-v.txt
then reboot the machine and attach /tmp/pg_ctl-status.out and
/tmp/rc_status-v.txt.