Mathijs Brands <mathijs@ilse.nl> writes:
> I've worked around the situation by running a small script that continually
> monitors postgres and takes appropriate action if postgres shuts down. I'm
> assuming this problem won't lead to any data corruption.
Hm. The problem here is that when the postmaster crashes, it probably
doesn't take down the old backends with it. So if you auto-restart
the postmaster, you may have two sets of backends that don't know about
each other. That would be A Bad Thing.
If your script kills off all the backends it can find before starting
the new postmaster, that should work OK.
A better solution would be to back-patch my errno fix into 7.0.3.
You can see the diff as it stands for current sources at
http://www.postgresql.org/cgi/cvsweb.cgi/pgsql/src/backend/postmaster/postmaster.c.diff?r1=1.198&r2=1.199&f=c
The additions to reaper() are probably the only part you really need.
regards, tom lane