Dedicating specific NICs to portgroups while maintaining NIC teaming and failover for the vSwitch

Products

• VMware vSphere ESXi

Issue/Introduction

The following describes ESX supported NIC team policy:

• ESX supports the combination of multiple physical network adapters (vmnics) into NIC teams.
• NIC teaming allows ESX hosts to take advantage of the additional network connection for additional network sessions
  and utilize the additional throughput for increased redundancy, fail over, and performance.

For more information about NIC teaming and Ether-Channel, see Sample configuration of EtherChannel / Link aggregation with ESX 3.x and Cisco /HP switches (1004048).

By default, ESX host directs network traffic for all defined portgroups out every available vmnic in the team. For example:

• One virtual switch, vSwitch0, with three teamed physical NICs, vmnic0, vmnic1, and vmnic2
• One Service Console portgroup, named Service Console
• One Virtual Machine Network portgroup, named Production
• One VMkernel portgroup, named VMotion

In this example, there are three portgroups in vSwitch0. Based on the default NIC teaming policy, all three portgroups may transmit and receive out of any available interface. As a result, the network traffic generated by one portgroup may affect the performance of another. In the sample configuration above, it is possible that high-volume VMotion activity may adversely affect communications on the other portgroups. The adverse affect on communication is because the default NIC teaming policy does not provide for isolation of one portgroup from another.

Environment

VMware ESX 4.0.x
VMware ESX Server 3.0.x
VMware ESX Server 3.5.x

Resolution

To provide isolation between different portgroups while still reaping the benefits of NIC teaming and its reliable failover mechanisms, you can modify the default NIC teaming policy to provide each portgroup with its own dedicated active network adapter.

To accomplish NIC teaming:

Note: These instructions use the example configuration from above.

1. Open the VMware Infrastructure Client and connect to the ESX host (or VirtualCenter) that you want to configure.
2. Select the ESX host object in the inventory and click the Configuration tab.
3. Click the Networking option under Hardware.
4. Locate the virtual switch containing the portgroups you want to modify. In this case, it is vSwitch0.
5. Verify that all three physical network adapters, vmnic0, vmnic1, and vmnic2, are assigned to vSwitch0.
6. Click Properties. The Properties window for vSwitch0 displays.
7. Click the first object in the list, vSwitch, and click Edit.
8. In the window that opens, click the NIC Teaming tab and verify that all three vmnics are displayed under Active Adapters.
9. Click OK to close the vSwitch Properties window.
10. Click the Service Console portgroup and click Edit.
11. In the window that opens, click the NIC Teaming tab and verify that all three vmnics are displayed under Active Adapters.
    You want the Service Console to be dedicated to its own vmnic, vmnic0, under normal circumstances. To achieve this, select vmnic1 and click Move Down to move vmnic1 into a Standby Adapter role.
12. Select vmnic2 and click Move Down to move it to the list of Standby Adapters. At this point, vmnic0 is dedicated to the Service Console, but if vmnic0 fails, the Service Console will fail over to vmnic1 and then vmnic2.
13. Click OK.
14. Click the Production portgroup and follow steps 11 through 14 to move vmnic0 and vmnic2 to Standby Adapters. This dedicates vmnic1 to the Virtual Machine Network portgroup unless vmnic1 fails.
15. Click the VMotion portgroup and follow steps 11 through 14 to move vmnic0 and vmnic1 to Standby Adapters. This dedicates vmnic2 to the VMotion portgroup while still allowing failover to the vmnic0 and vmnic1 adapters if vmnic2 were to fail.

When configured, this solution provides each portgroup with its own dedicated network adapter, effectively isolating it from the network activity of the others. However, it also allows each portgroup to be failed over to the two remaining network adapters if its own network adapter loses connectivity.