SAM-5 definition of physical reconfiguration

Knight, Frederick Frederick.Knight at netapp.com
Wed Jan 7 06:30:14 PST 2015


Attachment #1: <a href="http://www.t10.org/cgi-bin/ac.pl?t=r&f=r1501072_nameless-3396-2-1.html">nameless-3396-2-1.html</a>

And to add to Georges point.

Yes, the INQUIRY DATA HAS CHANGED is a must (since the VPD page data would
change); AND

An ASYMMETRIC ACCESS STATE CHANGED unit attention would also be appropriate,
since the REPORT TARGET PORT GROUPS returned parameter data would also have
changed.

Now, the question of course, about the ASYMMETRIC ACCESS STATE CHANGED unit
attention – is whether hosts assume that ONLY states changed (did the host
interpret this UA literally?), or if they interpret that unit attention in
the broader sense that “stuff” in the REPORT TARGET PORT GROUPS response
data has changed.

		Fred

From: Knight, Frederick

Sent: Wednesday, January 07, 2015 9:13 AM

To: 'Ballard, Curtis C (HP Storage)'; Ericson, George; Gerry Houlder; T10
Reflector

Subject: RE: SAM-5 definition of physical reconfiguration

I guess I tend to think of those words “physical  reconfiguration” as
meaning something happened that is “outside the scope of the standard”. 
I never considered whether it was S/W or H/W that cause the change just that
something outside what was documented in the standards had happened.

If it was just a MODE SELECT command that cause the change, then I’d expect
something in the standard to describe it.  Since there is nothing to describe
it, then I’d tend to say that a MODE SELECT doesn’t change that kind of
stuff.

		Fred

From: owner-t10 at t10.org<mailto:owner-t10 at t10.org> [mailto:owner-t10 at t10.org]
On Behalf Of Ballard, Curtis C (HP Storage)

Sent: Tuesday, January 06, 2015 5:54 PM

To: Ericson, George; Gerry Houlder; T10 Reflector

Subject: RE: SAM-5 definition of physical reconfiguration

I’m not certain that just removing ‘physical’ solves the problem but I
do agree that reconfiguration of virtual SCSI target devices can and should
be able to cause this.	We have devices that are able to be software
reconfigured so that virtual ports are enabled/disabled and we consider that
to ‘virtually’ be a physical reconfiguration.

The last time I looked at this we concluded that it took something
significant like a port add/delete to justify changing a relative port
identifier and that an external change like a cable pull/plug or a SCSI
change like an ADI port enable/disable command shouldn’t cause a change.

Curtis Ballard

Hewlett Packard

From: owner-t10 at t10.org<mailto:owner-t10 at t10.org> [mailto:owner-t10 at t10.org]
On Behalf Of Ericson, George

Sent: Tuesday, January 6, 2015 3:37 PM

To: Gerry Houlder; T10 Reflector

Subject: RE: SAM-5 definition of physical reconfiguration

The simple fix is to remove the word “physical” from the referenced
sentence.   Reconfiguration of “virtual” SCSI target devices has been the
practice for a very long time.	I expect an INQUIRY DATA HAS CHANGED event.

George

From: owner-t10 at t10.org<mailto:owner-t10 at t10.org> [mailto:owner-t10 at t10.org]
On Behalf Of Gerry Houlder

Sent: Tuesday, January 06, 2015 4:52 PM

To: T10 Reflector

Subject: Re: SAM-5 definition of physical reconfiguration

i tend to think of physical reconfiguration as a hardware change thing.
Changing cables around, moving a card from one slot to another, flicking a
switch. Things that are outside the scope of SCSI standards. i would consider
things that can be done with SCSI commands (e.g., MODE SELECT) to be more of
a logical reconfiguration that should't change a port identifier.

I will concede that we are now defining SCSI commands that can create or
remove logical units. This allows commands to create a virtual logical unit
in one location, tear it down, and recreate the same numbered logical unit in
another location within the same domain and have the same result as swapping
some cables or moving a PC card from one backpanel slot to another. This
muddies things a bit.

On Tue, Jan 6, 2015 at 3:10 PM, Kevin D Butt
<kdbutt at us.ibm.com> wrote:

In 4.6.5.2 of SAM-5r18 there is the following sentence.

The relative port identifier for a SCSI port shall not change once assigned
unless physical reconfiguration of the SCSI target device occurs.

What is the definition of "physical reconfiguration"?

Within my company I have opinions that oppose each other.

 1) physically change cables that are connected

 2) change the configuration settings (e.g., mode select)

Thanks,

Kevin D. Butt

SCSI Architect, Tape Firmware, T10 Standards

Data Protection & Retention

MS 6TYA, 9000 S. Rita Rd., Tucson, AZ 85744

Tel: 520-799-5280<tel:520-799-5280>

Fax: 520-799-2723<tel:520-799-2723> (T/L:321)

Email address: kdbutt at us.ibm.com

http://www-03.ibm.com/servers/storage/




More information about the T10 mailing list