Re: no DHCP-assigned InitiatorName

Julian Satran wrote:
> Michael,
>
> I will have to defer to boot RFC authors. My 2 cents is that DHCP
> "practice" has already several mechanisms to name the initiator (most
> based on what the DHCP agents present to the DHCP server - like the real
> of "fake" (for VMs) MAC address.

I agree that there are several different mechanisms that are used to
construct the initiator name. In practice, what we have ended up with is:

   iqn.1987-05.com.intel:<hostname>
   iqn.2000-01.org.etherboot:<hostname>
   iqn.1995-05.com.broadcom.<11.22.33.44.55.66>.iscsiboot
   iqn.1986-03.com.ibm.<11:22:33:44:55:66>.<hostname>

This is a problem because almost every commercially available iSCSI
target uses the initiator name as an identification mechanism to control
visibility to target LUNs.

As a result, trying to move from one iSCSI boot initiator to another
requires changes on the SAN storage controller (or iSCSI head) to
reconfigure for the new initiator name. This is unnecessarily
complicated and, in many commercial environments, forces coordination
across different organizational units.

> And I don't know how the iBFT interacts
> (or is supposed to) with a DHCP server.

There is no *direct* interaction between the DHCP server and the iBFT.
Rather, the values get passed through the iSCSI boot initiator.

The boot iSCSI initiator is responsible for populating the iBFT with the
iSCSI parameters that are required to continue to boot the OS once it
switches to its protected-mode drivers.

The relevant fields in the iBFT are:
  * portal hostname/ip addr
  * portal port
  * target name
  * target lun
  * initiator name
  * CHAP stuff

The iSCSI boot initiator fills in these fields regardless of whether the
parameters come from EEPROM config or from a DHCP server. The values
that are filled in are the same values used by the boot initiator for
its iSCSI login.


Michael
_______________________________________________
Ips mailing list
Ips@...
https://www.ietf.org/mailman/listinfo/ips