RE: an iSCSI name format for gateways?

["KRUEGER,MARJORIE (HP-Roseville,ex1)" <marjorie_krueger@hp.com>]

> I'd like to gage people's reaction to having a form of the iSCSI node name
> for gateways.

The iqn. format is sufficiently flexible to accomodate this need.  There's
no need to define another format, just define a format using iqn. that
gateway vendors agree to adhere to.

> Dave's presentation suggested that two separate gateways presenting 
> the same FC node could present a single node (with an "eui." name) with 2 
> portal groups.

And was told that this violated the rules of both SCSI and the iSCSI
standards, since this is the equivalent of two separate iSCSI devices
presenting the same name.

> An alternative suggested was the gateways should present the FC node as
two
> separate iSCSI nodes (so that each gateway would be an entirely separate
> I_T_nexus). 
> 
> One drawback of this alternative is that the configuration of the iSCSI
> network couldn't extend to the FC node but rather only to the gateway (for
> example, there'd be no way to have the FC node be a member of an iSNS
> discovery domain - only each individual gateway).

Why? (or why not?)  iSNS (or any other application) could certainly be
intelligent enough to understand a gateway naming convention and act
accordingly.

It's an important point that gateways that don't share information
pertaining to configuration and state are separate iSCSI devices.
Physically separate iSCSI boxes can only present the same name if they share
sufficient information to act as a single iSCSI device.