On Thu, 9 Oct 2003 nolan@celery.tssi.com wrote:
> > > It sounds like that is more a problem with improper operating protocols
> > > than with the underlying database.
> >
> > No. Problem is machine was shutdown with shutdown -h. It sends sigterm to
> > everybody. A good process would flsuh the buffers to disk before finishing.
> > Mysql didn't on that occasion.
>
> > Transactions or not, this behaviour is unacceptable for any serious app.
>
> True, but was it because the shutdown scripts weren't set up properly
> or does MySQL just not handle the 'kill' properly? (I would consider the
> latter a serious bug.)
>
> I still fault the operations protocol, part of what should be done in
> setting up a production shop is testing various shutdown options, and it
> sounds like that wasn't done in advance or they would have known to build
> in extra steps for shutting down MySQL.
>
> > Do a shutdown -h on a live database machine with pg. It will gracefully shut
> > itself down.
>
> Is that true for all OS flavors and is it dependent upon the DBA having
> set up proper shutdown scripts?
>
> I'm not trying to be argumentative here or defending MySQL, just noting
> that a shutdown process that isn't tested can cause problems even with
> commercial databases. And as someone who has to put up with MySQL on
> occasion, I'm always looking for problem areas for the DBA.
but a database should survive this kind of thing. Postgresql certainly
can survive having its power plug pulled mid transaction under heavy load.
that's what WAL is for. The fact that MySQL can be corrupted by being
shut down improperly points to a shortcoming a user should be aware of
before choosing it.
