Questions on relative target ports

wrstuden at wasabisystems.com wrstuden at wasabisystems.com
Tue Feb 3 16:26:43 PST 2004


* From the T10 Reflector (t10 at t10.org), posted by:
* wrstuden at wasabisystems.com
*
On Thu, 4 Mar 2004, Burn Alting wrote:

> * From the T10 Reflector (t10 at t10.org), posted by:
> * Burn Alting <burn at goldweb.com.au>
> *
> On Thu, 2004-03-04 at 03:21, Elliott, Robert (Server Storage) wrote:
> > * From the T10 Reflector (t10 at t10.org), posted by:
> > * "Elliott, Robert (Server Storage)" <elliott at hp.com>
> > *
> >
> > > Now my question is:- Is a 'relative target port identifier'
> > >
> > > a. with reference to the hardware (ie controller 0 port 0 == 0x01,
> > > controller 0 port 1 == 0x02, controller 1 port 0 == 0x03 and
> > > controller 1 port 1 == 0x03)
> > >
> > > OR
> > >
> > > b. with reference to the data set (ie if the data set is
> > > presented only on controller 1, port 1, then the relative target
> > > port identifier will be 0x01; alternately if it where on both
> > > host ports on controller 1, then port 0 would be 0x00 and
> > > port 1 would be 0x01).
> > >
> > > Thanks in advance
> > > --
> > > Burn Alting <burn at goldweb.com.au>
> >
> > I'm not sure what you mean by "data set of disks."  SCSI defines
> > logical units, target ports, and target devices.
> >
> > If there are 4 target ports providing access to the logical unit
> > (they're all in the same target device), they could be numbered:
> > controller 0 port 0: relative port identifier 1
> > controller 0 port 1: relative port identifier 2
> > controller 1 port 0: relative port identifier 3
> > controller 1 port 1: relative port identifier 4
> >
> > If some logical units are only accessible through some target
> > ports due to configuration (e.g. the Access Controls commands),
> > the relative port identifiers should still be consistent.
> > Relative port 1 should refer to the same target port, regardless
> > of which logical unit is being accessed.
> >
> > If certain logical units are only accessible through certain
> > target ports forever, then they might need to be considered
> > as being in two separate SCSI target devices.
> >
> > --
> > Rob Elliott, elliott at hp.com
> > Hewlett-Packard Industry Standard Server Storage Advanced Technology
> > https://ecardfile.com/id/RobElliott
> >
>
> Robert,
>
> Thanks for your e-mails.
>
> Sorry for the poor terminology, what I meant by 'data set of disks' is a
> set of disks which can be presented as a logical unit on one or more
> physical ports which could have differing scsi id's and scsi luns.
>
> Let's say I have two logical units A and B and my two port, dual
> controller raid setup. Unit A is presented on
>
> Controller 0, port 0	as SCSI ID 0, LUN 0
> Controller 0, port 1	as SCSI ID 1, LUN 0
> Controller 1, port 0	as SCSI ID 0, LUN 0
> Controller 1, port 1	as SCSI ID 2, LUN 0
>
> Unit B is presented on
>
> Controller 0, port 0	as SCSI ID 0, LUN 1
> Controller 1, port 1	as SCSI ID 2, LUN 1
>
> For logical unit A, the relative port numbers would be 1, 2, 3 and 4 - a
> "quad-ported" device.
>
> My question was essentially, what do I do about logical unit B. Since it
> is only ever "dual-ported", should it only have relative port numbers as
> 1 and 2, even though another logical unit presents different portnumbers
> on the same physical bus although a different lun?

>From my read on Robert's comments, since Controller 0, port 0 and
Controller 1, port 1 support different volumes, they should have
consistent rel port IDs across these volumes. If you had a Controller 0,
port 2 which only had a Unit C (Unit C was never on Controller 1 nor
Controller 0 ports 0 or 1, and Units A and B were never on Controller 0,
port 2), then its relative port IDs could repeat.

In other words, all the LUs in a given port, even for different units,
have to have the same relative port ID. Assuming I read Robert correctly.
:-)

Take care,

Bill
*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org




More information about the T10 mailing list