FCP-4: Letter Ballot Comment HPQ-219
Bob.Nixon at emulex.com
Bob.Nixon at emulex.com
Thu Oct 28 14:16:33 PDT 2010
Formatted message: <a href="http://www.t10.org/cgi-bin/ac.pl?t=r&f=r1010284_f.htm">HTML-formatted message</a>
In the first added paragraph, "...the Platform Name shall be the same as the
SCSI device name" raises a good idea, but adds a new "shall" too late in the
game for my taste. Platform Name has never been mentioned in FCP before. I
would be happy with a "should".
The long discussion of names reported by virtualized OSs leaves me confused
on a couple points:
1. It specifically relates to names "reported through all the SCSI
initiator ports". A Target has a VPD page, but how does an initiator report
2. Is there a change being suggested for FCP-4?
From: owner-t10 at t10.org [mailto:owner-t10 at t10.org] On Behalf Of David
Sent: Thursday, October 28, 2010 1:18 PM
To: t10 at t10.org
Subject: FCP-4: Letter Ballot Comment HPQ-219
At 9.93 in down and 0.41 in over
"Each FCP device should include a SCSI device name in NAA IEEE Registered
format (see SPC-4). If the FCP device includes a Platform Name (see FC-GS-6),
then the Platform Name shall be the same as the SCSI device name.
In the Device Identification VPD page, a device server in an FCP target
device that implements a SCSI device name:
a) shall report the SCSI device name in binary NAA format; and
b) should report the SCSI device name in SCSI name string format (e.g.,
"naa." followed by 16 hexadecimal digits followed by 4 ASCII null
Also add this to the SAM-5 names & identifiers annex (IEEE Registered format,
SAM-4 allows a transport protocol to mandate implementing device names and
define their format.
Node names were never well defined in FC, always unclear whether they named a
Port, an HBA (a set of Ports on the same card), or a system (set of cards in
a system). They are thus worthless.
Platform name supposedly provides clearer guidance, identifying the entire
system - the same scope as a SCSI device name.
With NPIV and server virtualization gaining popularity, it would be helpful
to have a unique identifier for each operating system instance, reported
through all the SCSI initiator ports (whether NPIV or physical) that the
operating system uses. If the operating system instance is shut down and
restarted on a different physical machine, that identifier should move with
it. This identifier should even work if the operating system has access to a
mix of protocols - e.g. some FCP ports, some iSCSI ports, and some SAS ports.
The same NAA IEEE Registered identifier can be reported and used in FCP (both
binary and as a "naa." string) , SAS (both binary and as a "naa." string) and
iSCSI (as a "naa." string). A system that doesn't have iSCSI ports could
just report the binary NAA format.
The device name would be helpful for configuring V-SANs, zoning, SCSI access
controls, etc. For example, the system administrator could grant certain
zoning permissions to an operating system instance, no matter which physical
machine it happens to be running on and which ports it happens to be using.
I have no problem with adding accepting this comment. Please respond with any
objections along with your reasoning asap.
Note I plan to move to accept the letter ballot comments at November T10 and
will have the comments and draft standard uploaded shortly. There will be one
comment open from my perspective and that comment is on the CAP agenda.
Also please consider the need/desire for a second letter ballot on FCP-4.
More information about the T10