Re: Shutting down a warm standby database in

On Thu, Nov 30, 2006 at 09:58:52AM -0500, Bruce Momjian wrote:
> Tom Lane wrote:
> > Bruce Momjian <bruce@momjian.us> writes:
> > > Stephen Harris wrote:
> > >> Should this cause a coredump when it happens?
> > 
> > > You should never get a core file.
> > 
> > elog(PANIC) => abort() => core dump.  This is completely expected.
> 
> Well, then I should have said there is no reason you should ever get a
> panic.

Well, as Tom said earlier in the thread
   > I see different results.  This time recovery aborts with a PANIC.      Yeah, that's expected since the whole
recoveryprocess is a critical   section.  We could change that but it's cosmetic.
 

Because of the changes made, we don't actually need to do a database
shutdown.  Merely killing the restore_command process (eg with kill -9)
appears to cause the same effect.

And now a personal opinion...

I think this is more than cosmetic; shutting down a standby database
cleanly is critical functionality for proper warm-standby procedures.
What we have now "works", but should be tidied up.  Probably low on the
priority queue though :-)

-- 

rgds
Stephen