Can't get NIC Teaming to work

[maaboo]

I'd like to set up new iSCSI storage for my hosts. I've chosen FreeNAS and installed it as VM on on of my ESXi hosts (I'm going to create few more VMs here for administrative purposes).

I've created two interfaces on FreeNAS (if I understand correctly it must be done to use IP Hash policy):

When I create load on FreeNAS it shows traffic on both interfaces:

Also I see traffic on both paths at initiator host:

I've made aggregated ports and set up the ESXi host as described here:

(all port groups have the same settings)

But I get only one NIC active on FreeNAS'es host (what is working on less than 50% speed):

What have I missed or misunderstood?

Any ideas will be much appreciated. Thanks in advance.

[DavoudTeimouri]

Did you configure link aggregation on those ports?

[RiaanBadenhorst]

You've got two hosts that we can see, 241 and 243. On 243 the avg speed seems to be the total of both cards on 241. You're using standard vswitches, so have you configured the teaming on the other switch (243)?

[hussainbte]

IP Hash chooses a physical NIC uplink based on a hash of the source and destination IP addresses for each packet. If a single IP address pair exists for the source and destination of all traffic, the traffic will only utilize a single physical uplink.

Check the linked KB

NFS with IP Hash Load Balancing (1006795) | VMware KB

[RiaanBadenhorst]

Hi,

But there are two IP address pairs as the Source has 2 addresses. Or does it not work that way?

[hussainbte]

No It Doesn't work that way.

Once the Protocol chooses a particular nic based on a hash value it continues to use it as long as the nic is active.

Also, having 2 physical nics and one being already used for a hash value does not mean that the same nic cannot be used for another hash.

I don't think we can use nic teaming with IP based storage.

Check the below link.

Why can you not use NIC Teaming with iSCSI Binding? - VMware vSphere Blog

[RiaanBadenhorst]

Hi Hussain

Thanks. Interesting topic in the link.