Splitting Etherchannel ports between multiple vSwitches?

[PredatorVI]

I'm hoping for some enlightenment since my current understanding is lacking and I can't find the answer elsewhere.

My ESXi Servers have 8 x 1GbE ports that we use with (currently) two VLAN ids (110, 124) connected to a Cisco 3750 switch stack.

The configuration has two Etherchannel port groups configured on the switch stack.  Each port group is set to use 802.1Q trunking and Mode set to "On (No LACP)" on the Cisco switches with 4 ports assigned to each.  These directly map to individual vSwitches as shown below.  The vSwitch load balancing algorithm is "Route based on IP Hash" and all seems to work fine in this scenario.

However we ran out of Etherchannel port group IDs (48 Max) on the switch stack.  I thought that I could simply combine the Etherchannels and connect all 8 physical ports into ONE 8-port Etherchannel and keep the existing dual vSwitch configuration to maintain functionality.  However that didn't appear to work.

If 4 ports are assigned to vSwitch0 and 4 to vSwitch1 as shown above, with a single, 8-port Etherchannel configured at the switch, communication stopped working.

We were able to get it to work if we combined the two vSwitches into one such as:

My question is why didn't it work with two vSwitches?   There is probably a simple explanation but I'm lacking some understanding.

Thanks!

[parmarr]

As you see an issue with multiple vSwitches configuration, Click on :smileyinfo: to Show details about the configuration to confirm the status like connected, enabled VLANs, duplex, available networks etc etc. Configuring multiple vSwitches with same configuration shoudn't be a problem until you missed some configuration steps or assigned incorrect Physical adapters(vLAN) to those vSwitches.

[Finikiez · Accepted solution]

I guess the answer is simple.

Your Cisco switch is not aware of vswitches and port groups on host

In configuration with single etherchannel and 2 vswitches the following scenario is possible: VM in port group with VLAN 124 can send traffic throug vmnic0-3, but it can happen that ESXi host can get reply on vmnics 6-9. Because VM is not connected to vSwitch1 - network packet is dropped by host.

The same happens with vmkernel interface as well.

[PredatorVI]

Thanks!  I thought that when communication happened on one interface, the switch would route responses back through the same.  Maybe I'm confusing the non-Etherchannel configuration.  Regardless, this makes perfect sense.