ESXi NIC Redundancy

[davequinlisk]

Hi All,

I have configured an ESXi host with two physical NICs within a team.

If I was to remove one of the physical NIC's from the team when using the vSphere client, should I lose connectivity?

I have the following configuration:

HP c7000 enclosure

Cisco 3125 stacked switches in IC bay 1 + 2

HP BL490c ESXi 4.0 Update 1 host.

If I 'shutdown' the switch ports on the Switch, failover is succesful but if I manually remove a NIC from the team within the vSphere client the connection drops.  I would expect this just to fail over to the second NIC in the team.  Could someone validate what the expected behaviour should be.

Many Thanks

Dave

[bulletprooffool]

When you say remove nic, are you unpluggun the network cable?

If you right click on the vSwitch you have created, you can view the vSwitches properties. you will need to amend your Load Balancing  and Network failure Detection settings in order to streamline the failover process.

You need to make sure adapters are set to Active / Active, not unused etc.

Good luck

[davequinlisk]

Hi

Thanks for the response.

I do not mean removing the network cable.  If you point your vSphere client at your ESXi host and navigate to the following >

<Host> | Configuration | Networking | vSwitch0 Properties | Network Adapters

Within this section I have two NICs configured with the Load balancing policy > 'Route based on IP hash'.  (in-line with KB 1004048)

If you remove 1 of the NICs from this team I am losing connectivity.  Is this expected behavior?  do you require two NIC's within the team for the team to function?

Thanks

Dave

[AndreTheGiant]

Usually you may temporally lost the connectivy (if the connection was on those link).

But in your case you have an etherchannel so probably do not loose the connection.

Note that the NIC is connected to a switch port "bonded" with another port.

So be careful before re-assign it to another vSwitch, check and fix the physical switch configuration.

Andre

[Josh26]

Hi,

If we assume your switch is setup with an appropriate team - and that team is pulled apart from the ESXi configuration, but left in place on the switch - as appears to be your process - it would be expected that this behaviour would occur.

[davequinlisk]

Thanks for the responses so far!  I just needed to validate this should work as I expect.  In my current test secnario I am not experiencing expected behaviour.

I have completed some further testing today and added to my confusion rather than progressed!

If you take a scenario where you have have one ESXi server and two separate clients all running a 'ping <IP Address> -t'

When I remove 1 of the NIC's from the ESXi host team then 1 of the two PC's will lose their connection to ESXi and the other will maintain its connection.

If I reverse the conditions and re-add that orginal NIC back in and remove the other NIC the remaining PC will lose it connection and the original PC will come back online.

(Just to add, if I shutdown either port on the switch itself then this fails over as expected, this is only an issue when removing the NIC within the ESXi configuration screens.)

Definitely seems to be a Load balancing issue, any help on solution or troubleshooting steps are appreciated!

Thanks

Dave

[CSHAW78]

You need to enable notify switches ..

With this enabled the switch is able to detect that the physical NIC has failed and a failover has occured to the second NIC in the team.

[davequinlisk]

Thanks for the message CSHAW.

Notify Switches is already enabled..  I agree it does sound like an issue that could have been related to this.  Whether Notify Switches is on or off it fails.  If anyone could provide lower level detail of how the Notify Switches function works I can investigate this on the Cisco side, what protocol does it use to notify the switch?

Thanks

[Court01]

I agree with Josh26.  The physical switch doesn't know that you removed the pNIC from the vSwitch, it is still seeing it as connected.  Therefore it is still trying to load balance traffic over both interfaces in the port channel.  Have you tried pulling the network cable.  This would cause the Physical switch and the vSwitch to see that there was failure of some type and divert all traffic over the other active pNIC.

[davequinlisk]

Hi Court01,

Thanks for the response.

As mentioned in my previous post, I have tried shutting down one of the ports from the Cisco switch and everything works as expected.

I am just surprised at the behaviour of when you remove a switch from the ESXi configuration.  I would have expected a scenario that if you remove a single NIC from the team, the switch would be notified and the alternative link would be used.  This would be useful for a scenario of modifiying your NIC configuration without having to go into Maintenance Mode.

As it seems this is expected behaviour then question answered.

Thanks for the responses!!

Dave