--xgyAXRrhYN0wYx8y
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Sun, Apr 11, 2004 at 06:50:48PM +0900, Curt Sampson wrote:
> On Fri, 9 Apr 2004, Bill Studenmund wrote:
>=20
> > A simple example is a failover
> > configuration, where the hot-spare computer can grab the disk on failur=
e.
>=20
> That is not quite so simple as it seems, actually, because for this to
> work reliably (as opposed to trash the entire disk :-)) the takover host
> needs to have some way to ensure that the allegedly failed host cannot,
> even if it wants to, use that disk any more. Basically, the host taking
> over the device needs some sort of "cutoff switch" under its control that
> cuts off the other host from the system in question.
>=20
> The issue is, of course, that the allegedly malfunctioning host may seem
> to be down, but may not really be down (or down enough), and may still
> try to use that device.

Check out RESERVE, RELEASE (both now obsolete), PERSISTENT RESERVE OUT,=20
and PERSISTENT RESERVE IN. SCSI has primitives, these commands and the=20
associated reservation system), that handle the problem you describe.

Take care,

Bill

--xgyAXRrhYN0wYx8y
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (NetBSD)

iD8DBQFAeZ7tWz+3JHUci9cRAiVhAJ4pw3XD6tc5/PWLbskd6BvtRZvU3wCfXh7r
+nD1NX/YCHT5rhtgt+xpLjI=
=5njK
-----END PGP SIGNATURE-----

--xgyAXRrhYN0wYx8y--