One thing about the ISCSI initiator in 3.x was even with 2x adapters bound to the vSwitch, the iSCSI adapter would only ever show a single path to the lun, and the only time it would use the secondary adapter is if the primary one failed.
Now something I noticed in the new vSphere iSCSI SAN configuration guide was now we are able to bind 2 network adapters via a single vSwitch with two port groups, or bind 2 x vSwitches with a single adapter and port group each to the built in software iSCSI initiator.
I started up my vSphere lab system and took some screenshots and thought id share a few of the key steps. Ill be honest I’m not great at sitting down and following documents step by step so of course I got an error which ill also explain and show what was doing wrong.
So here you ll see Ive started off with only vSwitch0 with the service console attached.
Next I goto Add Networking and choose to add VMKernel and then add vmnic1 and vmnic2 to the new vSwitch.
Now that I have both adapters bound to vSwitch1 I need to go and make a small change and set the secondry adapter for each port group as “unused”. If you dont do this then later on in the section where we add both port groups to the iSCSI Initiator you ll get the following error “Add Nic Failed in IMA”
The next step is for me to select the first port group which I called iSCSI_VMKernel1 and then select the NIC TEAMING tab. Then Select the override option and move the secondary adapter to unused, note for the next port group it will be the other way around.
So now Im going to go and change this for iSCSI_VMkernel2. Note this time its vmnic1 that is set to unused.
So that’s the configuration of the vSwitch and port groups done. Now lets bind both vmk0 and vmk1 to the iSCSI Initiator using esxcli.
esxcli swiscsi nic add -n vmk0 -d vmhba33
esxcli swiscsi nic add -n vmk1 -d vmhba33
Now I want to confirm that everything worked as expected.
esxcli swiscsi nic list -d vmhba33
Yep thats all looking good.
Now lets connect my vSphere server to the iSCSI Target.
Now lets take a look after selecting YES to a rescan.
Great now we have two paths to the lun.
Now how much better is that then ESX 3.x, Oh and dont forget to go and change the path policy to “Round Robin” which makes both path active.