Software iSCSI Adapter - Network Port Binding

Varez IT

I'm currently trying to bind a VMkernel adapter on a standard vSwitch in VMware ESXi 5.5 to a software iSCSI adapter but for some reason it will not work. It is getting late here so I may be brain fried but here is what I have done.

1. Created a VMkernel adapter on the VMware ESXi 5.5 host.
2. Enabled the software iSCSI Initiator on the VMware ESXi 5.5 host.
2. When I click the green plus icon to add a VMkernel adapter to the software iSCSI initiator the list is empty.

What am I missing?

Thank you.

Deathmage

did you add the static/dynamic address of the IQN?

Essendon

Follow this article.

Do your hosts see the iSCSI array at all?

joelsfood

Your standard switch has two or more uplinks. You need to override the adapter binding on that VMK port and bind it to just one uplink adapter.

Varez IT

Deathmage, yes, the IQN has been added to the software iSCSI initiator.

Essendon, what a great article! Thank you for providing that but I have followed those steps with the exception that I have added (2) vmnics to the vSwitch.

With that being said, I believe that joelsfood is correct. I won't be able to totally confirm this until I get home later.

This configuration is for a lab and I have configured the software iSCSI initiator on multiple occasions so I was confused why it wasn't working last night. I do remember that last time I only had (1) vmnic bound to the vSwitch. I was also brain fried last night after hours of study so I didn't even think of that. I'm also going to change the configuration and use a dvSwtich (after I confirm the vSwitch with a single vmnic).

You gentlemen have been very helpful! I'll let you know what I find tonight.

Deathmage

it might be possible that nothing changed on the VMware side. If this is a home-lab is your iSCSI network on a physical switch and/or NAS/SAN? - if this is a yes, I know it happens to me, check the obvious like cables and the iSCSI switch make sure the configs are correct, switches powered-on, and cable connected properly it might be a layer 1 issue and not a layer 7.

if it's a NAS or similar device make sure permissions are working correctly and folder/lun shares are accessible from your storage device. I'd simply plug in your laptop and ping all of the IP's I could on the iSCSI fabric to verify it's not a layer 1-2 error then you can with a certain level of confidence remove networking as the cause.

Varez IT

Deathmage,

Thank you for providing that information but this is a nested virtualization configuration with VMware Workstation 10.

Deathmage

hmmm... I never did a nested, I went the complete lab approach.

Varez IT

Deathmage,

Yes, I would have preferred the complete lab approach but I went with nested.

Essendon

Okay then, check this:

- the vmk adapter you added, did you associate one and only one uplink to it? If you havent, give it one active uplink and move the others to unused.
- does the switch have uplinks at all?

Basically, just elaborated a little on what Joel said.

Varez IT

Essendon,

Originally, the VMkernel adapter had two uplinks associated with it. I reconfigured it with one active and the rest unused and that resolved the issue. I was able to test that configuration with a vSwitch and dvSwitch.

Although this was a simple fix it was certainly a good learning experience.

Thank you all for your assistance!

Deathmage

Varez IT wrote: »

Essendon,

Originally, the VMkernel adapter had two uplinks associated with it. I reconfigured it with one active and the rest unused and that resolved the issue. I was able to test that configuration with a vSwitch and dvSwitch.

Although this was a simple fix it was certainly a good learning experience.

Thank you all for your assistance!

Two or more uplinks would only work in IP hash mode for a bonded/teamed connection. If you used two uplinks at the same time you essential cause a broadcast storm since I'm pretty sure in a nested world you can't enable Spanning-Tree Protocol. If it was a physical connection on say a Cisco switch I bet it would have worked because STP would have kicked in and blocked one of the port automatically. A two uplink scenario is good for fault tolerance but without STP it can cause problems.

I'm sure some would correct me if I'm wrong but from a networking perspective that's what it sounds like happened.

joelsfood

Not an issue of STP (multiple uplinks without LACP, the host uses mac pinning to bind flows to uplinks, avoiding mac flapping, etc). It's actually a limitation of the software iSCSI adapter that it will only let you bind to vmk with a single uplink.