Multiported disks

Paul von Behren Paul.Vonbehren at sun.com
Mon Jun 27 09:25:27 PDT 2005


* From the T10 Reflector (t10 at t10.org), posted by:
* Paul von Behren <Paul.Vonbehren at Sun.COM>
*


Burn Alting wrote:

>* From the T10 Reflector (t10 at t10.org), posted by:
>* Burn Alting <burn at goldweb.com.au>
>*
>Hi,
>
>I have two question about multiported logical units (eg disks).
>
>If a logical unit is multiported, then should each of say the T10, EUI64
>or NAA identification descriptors be different depending on which port
>or 'path' the logical unit is accessed via?
>  
>
>For example, lets say I have a dual ported fibre channel disk.  I
>connect one port to one HBA and the other port to another HBA and issue
>INQUIRY commands to the disk requesting the device identifier VPD (page
>0x83) via both HBA's. Should, say the reported T10 or EUI64
>identification descriptors be identical or different?
>  
>

SCSI VPD page 83 can return a list of ID descriptors.  The descriptors 
include a field called "association"; association = 0 indicates the ID 
is associated to a logical unit, association = 1 indicates the ID is 
associated with the target port, and (in SPC 3) association = 2 
indicates the ID is associated with the target (independent of 
ports/LUs).    I'd expect the association=0 IDs to be the same across 
target ports.


In Solaris 10 (and 9 with appropriate packages), we expect multi-ported 
disks to return the same association=0 ID regardless of target port.  
That ID along with the target port group information allows drivers to 
automatically treat the disk as multipath and provide generic failover 
capabilities without requiring device-specific drivers.


>My second question relates to VPD page code 0x80 (serial number) and a
>multi ported logical unit. Should the serial number for a logical unit
>be the same regardless of the access path? That is, is the serial number
>of a logical unit bound to the logical unit or the logical unit and port
>it is accessed via?
>  
>

The page 80 documentation has permitted some wiggle room - and in 
practice, devices seem to have wiggled a lot. The docuemntation in SPC 
for page 83 is much more depcific.  For that reason, Solaris prefers 
page 83 IDs over page 80 IDs; page 80 IDs are used as a fall-back in 
earlier versions of Solaris and I believe the current versions don't use 
page 80 at all. 

Page 80 does not have the equivalent of page 83's "association".  In my 
observations, devices expose the same page 80 ID regardless of port.  
Unfortunately, some multi-LU arrays report the page 80 ID for all 
logical units. 

Paul

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