Re: iSCSI: 2.2.6. Naming & mapping

["John Hufferd/San Jose/IBM" <hufferd@us.ibm.com>]

Peter,
As with Fibre Channel (and even multiple SCSI buss connections), once you
have acuired a connection to the Storage Controller, you may address  LU 0,
to get the LU numbers that are authorize to that connection (with some
Storage Controllers these are virtual Devices authorized by connection
"port" and/or WWN of the initiator).  The values returned by the LU0 query
are the unique LUNs related to that LU 0  (real or virtual).

The Host SCSI layer can then obtain a unique ID from the EVPD page 83h.
When the same unique ID is found on different paths then they have
discovered an "Alternate Path".   All of this is normal SCSI, not
transport.

Since many Storage Controller create virtual volumes, and these are only
visable to authorized Initiators,  and also unknown to the transport. It
seems that we should keep to the same approach used by Fibre Channel et.
al. and keep it away from the (iSCSI) transport.

Here is what Jim Hafner said to the ips working group WAY BACK on 6/19/00:

{snip}
SCSI HAS defined a "WWN associated with a volume" with EVPD page 83h.
This is independent of LUN value and also independent of the transport
layer.  You (should) get the same answer if your logical unit
is connected (simultaneously) to FC port and to a parallel bus.  And use
of this WWN is already supported as a naming specification in other places
in SCSI .....
{snip}

.
.
.
John L. Hufferd


Peter Johansson <PJohansson@ACM.org>@ece.cmu.edu on 09/16/2000 04:45:23 PM

Sent by:  owner-ips@ece.cmu.edu


To:   IP Storage <IPS@ece.cmu.edu>
cc:
Subject:  Re: iSCSI: 2.2.6. Naming & mapping



At 04:31 PM 9/14/00, John Hufferd/San Jose/IBM wrote:

>There is NO SUCH THING as an iSCSI LU.

I agree, John, that there is no such thing as an iSCSI LU. But that wasn't
the issue that was raised.

SCSI targets are addressed by an iSCSI method. If there is more than one
path by which an iSCSI target is addressable (perhaps a target with more
than one IP address or URL?), then you had better require unique IDs of
SCSI LUs. Otherwise it is not possible to discover that what appears to be
two LUs, reached by different paths, are in fact the same LU.

One may associate a unique identifier with an LU by either a) command set
methods (SCSI) or b) transport protocol methods (iSCSI in this case).

SBP-2 elected to make an LU's unique identifier visible at the transport
protocol layer. This was chosen to permit a host to enumerate unique LUs
(and identify redundant paths) prior to a login to use the device.

I'm not fully conversant with iSCSI device discovery methods, but it might
be desirable to make a unique LU identifier visible during this stage of
the process, for reasons analogous to those discovered by SBP-2.




Regards,

Peter Johansson

Congruent Software, Inc.
98 Colorado Avenue
Berkeley, CA  94707

(510) 527-3926
(510) 527-3856 FAX

PJohansson@ACM.org