Heikki Linnakangas <heikki.linnakangas@enterprisedb.com> writes:
> Craig Ringer wrote:
>> Is there any chance you have EVER hard-killed the postmaster manually
>> (eg with "kill -9" or "kill -KILL")?
> There is safeguards against that. If postmaster dies, the backends
> should die quickly and gracefully too. And postmaster refuses to restart
> until all the backends have died and detached from the shared memory
> segment.
A sufficiently bull-headed DBA can defeat those safeguards though ---
from memory, manually removing postmaster.pid and then starting a new
postmaster while old backends remain alive is the ticket to data
corruption.
regards, tom lane