Subject Re: [IB-Architect] GBAK processing
Author Jim Starkey
At 12:09 PM 4/26/00 -0700, Bill Karwin wrote:
>----- Original Message -----
>From: "Jim Starkey" <jas@...>
>> Would a better solution be a mechanism to create and
>> detach a shadowed copy? Unlike the gbak case it would
>> be a complete usable copy.
>The other problem is that shadows cannot currently write to NFS filesystems,
>I believe (no reason why they shouldn't be able to, but that's the way it is
>today). So if it's the drive or the server hardware that causes the
>downtime, you haven't achieved anything, unless you copy the shadow to
>another server after you detach it.

I suspect that derives from problem that NFS doesn't give any good
mechanism to enforce system affinity for database connections, and
has nothing to do with shadows. This obvious can and should be
changed, since shadowing across NFS should be the norm on Unix
systems, not the exception.

Creating a shadow set has not more performance implications that
creating a copy of a database with the added benefits that it
actually works.

I continue to think that the detachable shadow idea has a great
deal more merit than trying to make poor gbak (originally known
as "burp") do unnatural acts.

Jim Starkey