Connectivty lost after removing a vNic from a Portgroup

[Teovmy]

We are suffering a strange problem regarding networking within an ESX host.

To outline the problem we configured a vSwitch1, and added 4 vNics (vmnic2, vmnic3, vmnic4, vmnic5)

Further configurations on the vSwitch1:

Load Balancing > "Route based on IP hash"

Network Failover Detection > "Link status only"

Notify Switches > "No"

Failback > "Yes"

Security and Traffic shaping left default.

Second, added a Portgroup named: Production, and Tagged it with a vlan.

One virtual machine called VCS02 is running on this vSwitch1.

When we remove vmnic5 from vSwitch1 the VCS02 virtual machine suffers a ping lost. No connectivity is possible.

When we add vmnic5 to vSwitch1 the machine is reachable again.

What happens here? I guess it should be possible to remove a vNic without any consequences?

We also reconfigured all vswitches, portgroups etc. No luck here.

The physical port is configured similar as shown below

#int port-channel1
#switchport trunk encapsulation dot1q
#switchport trunk allowed vlan all
#switchport mode trunk

#int gi0/23

#switchport trunk encapsulation dot1q
#switchport trunk allowed vlan all
#switchport mode trunk

[MarekZdrojewski]

Hi there!

Is this a Cisco/HP switch? If so, please check this KB article on EtherChannel trunking. This has to be static not dynamic on the switch.

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=100404...

hth

Regards.

[rickardnobel]

Teovmy wrote:

When we remove vmnic5 from vSwitch1 the VCS02 virtual machine suffers a ping lost. No connectivity is possible.

When we add vmnic5 to vSwitch1 the machine is reachable again.

What happens here?

When you say "remove vmnic5 from vSwitch1", does that mean removing it in the configuration, by the vSphere Client? Or do you pull the cable?

If you are removing it logically from the vSwitch then the traffic loss is most likely caused by the physical switch still sees that the link is up and continues to forward frames to both links in the etherchannel, but on the Vmware side there is no vSwitch that accepts the packets.

[Teovmy]

Thanks for the help. In the meanwhile this is the sollution.

So the problem is here:

We have an etherchannel configuration on our physical switch. 4 switchports are the members where my vmnci2-5 connects.

On vSwitch side we have two vswitches where we VM connected with two interfaces.

If I connect all network card to vSwitch1, it works fine, however if I remove vmnic5 from vSwitch1 and add it to vSwitch2, connection will be lost.

This happens because on the physical switch side the etherchannel still using all four network card, but you broke the etherchannel settings on virtual side. The physical switch does not know anything about that you moved the vmnic5 to a different vSwitch, so it will send the traffic via that card to the wrong interface. This is the expected behaviour.

If you want to use this settings, you must remove the switchport from ether channel settings where vmnic5 connected. This will fix your problem.

[rickardnobel]

Teovmy wrote:

In the meanwhile this is the sollution.

Yes, as I wrote above this is the expected behavior. You could not have a link aggregation from one switch connection to more than one other switch, virtual or physical. If you want to have several vSwitches then you must create multiple Etherchannels on the physical switches.

[Teovmy]

Thanks for your help. I appreciate that.