[Fwd: 2 computers 1hd 2 postgres daemons. Is it possible?] - Mailing list pgsql-interfaces

From Joseph Shraibman
Subject [Fwd: 2 computers 1hd 2 postgres daemons. Is it possible?]
Date
Msg-id 39ECA044.40AA6EAB@selectacast.net
Whole thread Raw
List pgsql-interfaces
--
Joseph Shraibman
jks@selectacast.net
Increase signal to noise ratio.  http://www.targabot.comKris Klindworth wrote:
>
> > "Clark, Joel" wrote:
> > >
> > > Yeah, I was really just curious if the proposed system was at a system leve
> > l
> > > or a network level.  Having two physical machines share (at the system i/o
> > > level) a storage unit is something I have never heard of.  Does such a beas
> > t
> > > exist?
> >
> > Yeah, they exist, though why they exist is beyond me.  Maybe because
> > windows systems crash all the time.  It doesn't make much sense to me,
> > after all it would seem that hard drive failure happens more often that
> > the system crashing for no reason.
>
> It is a down time issue.  We can't afford to be down for hours at a time
> when a chip fails.  It doesn't happen often, but it does happen and when
> it does it is a VERY big deal.
>
> We have two identical boxes that provide two distinct service sets.
> The two machines and the drives for their services are on a shared
> SCSI bus.  In normal mode, the drives are mounted on their normal hosts
> and the systems monitor each other.  If one sees that the other has gone
> down, it will automatically mount the drives for the missing services,
> start up the applications, and we are back in business with zero manual
> intervention.  It just sends out an email warning us that the other
> system went down and that it had taken over the services.  When the
> repairs are completed on the downed box, we pick a convenient time to
> interrupt the services and then we move them back to their normal host.
>

Then you run into problems with both systems failing to detect the other
and thinking it is the only one left running, so you end up with two
servers running.

My point is that I can't conceive of a failure that is likely to happen
where a shared scsi bus makes a difference. If a hard drive goes down it
is no help. If the application software goes down it doesn't help
because they share the same application code.  I would think the added
complexity of a shared scsi bus would create more failures than would
happen on a normal system from a hardware chip failing or something.

But then again what do I know?


--
Joseph Shraibman
jks@selectacast.net
Increase signal to noise ratio.  http://www.targabot.com

pgsql-interfaces by date:

Previous
From: Dnesbitt@encryptix.com
Date:
Subject: JDBC support for DROP USER?
Next
From: Tom Lane
Date:
Subject: Re: 2 computers 1hd 2 postgres daemons. Is it possible?