Start a Conversation

Unsolved

This post is more than 5 years old

20103

September 22nd, 2010 23:00

Why does MD3220i need multiple subnets on the iSCSI ports?

My only experience with iSCSI is from using OpenFiler and ESXi 4.1 networking. I assigned one subnet to the openfiler's target and created one vmkernel switch with NIC teaming in my four esxi hosts. All the hosts saw all the LUNs and everything was fine. Now we have the MD3220i with the eight port dual controllers. I am pulling my hair. Dell tech support hasn't been helpful with my question.

Why do I need to assign multiple subnets to the eight ports on MD3220i? The guide says each port MUST be in separate subnets. Why can't I put all the iSCSI ports in a single subnet.

I have four R710 ESXi 4.1 hosts with four NICs for MD3220i. Four hosts connect to two powerconnect switches which, in turn, connect to MD's eight ports. If I follow the guide, there will be four subnets. Am I right in saying that I will need four vmkernel ports in each ESXi host?

I appreciate your help.

34 Posts

September 23rd, 2010 09:00

Typical 'best-practice' for iSCSI is separate subnets for the iSCSI targets... this ensures security, fault-tolerance, and helps the SAN provide multiple paths for MPIO. For the MD3220i, since you have 2 controllers with 4 ports each, I would think you would need 4 separate subnets.

847 Posts

September 23rd, 2010 10:00

On certain iSCSI controllers, basic network routing can get confused on the san side and performance can suffer because of it. Generally if the guide for the san shows different subnets you should adhear to it. On other san models the guide will say to use all in one subnet. It has to do with how robust the built in routing on the san is, I think anyways.
No Events found!

Top