Relation between IP address and iSCSI targets?

[rickardnobel]

I have some questions regarding iSCSI and IP.

1. The iSCSI initiator and the iSCSI target "emulates" an normal SCSI initiator and target and uses TCP/IP to carry the SCSI commands inside iSCSI PDUs over a typical Ethernet network. Is it correct logically that IP is only the carrier and that the Vmkernel iSCSI initator only "sees" paths (through one or several vmknic/vmnic bindings)?

.

2. Is the iSCSI Target logically equalient with a "FC Controller" as it presents a certain amount of LUNs to certain hosts?

.

3. If two iSCSI Targets are available behind the same IP, what does that really mean? That is, if two targets presents a certain LUN, we can have

multipath, but does this has any value if only one IP address available for the network communication?

Above exampel with ESXi host with two vmknic/vmnic bound with esxcli and a iSCSI server presents LUN 0 through two iSCSI targets with different IQN, but same IP.

Should two iSCSI Targets which exposes the same LUNs for multipathing purposes typically have two different IP addresses?

[RLX]

Howdy Ricnob,

I'll take a stab at this:

1. Yes, and, unless you have a NIC that does iSCSI offload your CPU is doing all that emulation for you so leave free processing power to do so.

2. Not certain what you mean by "FC Controller."  I'm assuming you're referring to FC zones used to manage FC hba visibility.  You can use an iSNS management server/device to do this in an iSCSI environment.  Alternately, most storage devices allow whitelisting as an access limiter.

3. The answer to this is: it depends.

It depends on the storage device you're using, and the configuration of said device.  With only one IP available I'm guessing the SAN is configured to handle nic failure on the SAN side.  So while you see only one IP there's actually multiple nics in the storage array in an active/active configuration answering to that IP.  The way this is done varies based on network and storage hardware.

tl;dr - this is really a question for your SAN and/or network administrator(s).  If you are only talking to one nic on the storage array then your VMWare MPIO is wasted.

Note that an active/passive setup on the storage side = talking to one nic.  If that's the case you'll want to change to a design that allows the use of the second nic (i.e. active/active or a second IP.)

Direct your SAN admin to http://www.vmware.com/pdf/vi3_35/esx_3/r35/vi3_35_25_iscsi_san_cfg.pdf for 3.5 or http://pubs.vmware.com/vsp40_i/wwhelp/wwhimpl/js/html/wwhelp.htm#href=iscsi_san_config/c_using_esx_e... for 4.0