Switch Port configuration for ESX host

[TheVMinator]

I have a switch which will connect to a physical server used in

virtualization. This means that virtual machines with IP addresses on

multiple subnets will be sending layer 2 frames to this switch.

However, in attempting to minimize the number of NICs on this server

there is a question. If there were only one NIC in this physical

server connected to one port on the physical switch - is it possible to

have one port on the physical switch handle frames which originate from

all VMs running on the physical server - when these VMs have IPs on

different subnets? Or does each separate logical IP subnet require a

separate switch port to handle its frames and a separate physical NIC on

the server used for virtualization?

Thanks for your input

[jamesbowling · Accepted solution]

Yes, you would need to configure the physical switch port as a trunk and then configure VLAN tagging on your portgroups in Configuration->Networking.

If you found this at all helpful please award points by using the correct or helpful buttons! Thanks!

[golddiggie]

You could do it with one pNIC on one vSwitch with different port groups, but it's really going against BKM's and recommended configurations up the whazoo...

A more typical configuration (ruling out any needs for storage, such as a SAN/NAS using either NFS or iSCSI) would dictate at least two vSwitches with a minimum of one pNIC per vSwitch (BKM's want dual pNIC's per vSwitch for redundancy and load balance)... Use vSwitch0 for the management network (ONLY!) and then vSwitch1 for your Virtual Machine traffic (VM Port Group)... If you need/want, you can mix the pNIC settings up a bit having one port group use pNIC0 as active with pNIC1 as standby, then reverse that to have both pNIC's in active use all the time... A fully redundant configuration would also have each vSwitch span two physical switches, with the physical switches using interconnects to communicate (such as the 10Gb interconnects present on switches like the HP ProCurve lineup)...

Unless you have very, very, very few free physical switch ports available to you, I would go with at least four (six if you're using a network connected SAN/NAS) and have a proper configuration... Using just one means your network connections could have collisions, reducing the speeds, causing pain points that are easily avoided... Even in a test/lab environment, I would follow this method... Actually, I pretty much am in my home/test lab...

VMware VCP4

Consider awarding points for "helpful" and/or "correct" answers.

[TheVMinator]

Thanks for the input. Actually, the issue really isn't not having enough NICS - I was just picturing the one physical NIC scenario to simplify the real question which has to do with IP subnets. If VMs from multiple, differing IP subnets need to be on one ESX host, and there are say 20 of these different subnets but 8 physcial NICS - the question is really about traffic originating from VMs on more than one subnet can travel on the same physical NIC to the same physical switch port and the physical switch be able to process all frames without issues. So creating a hypothetical scnerario with one pNIC (although a false scenario) was more to simplify question of traffic from more than one logical subnet going on a single physical NIC.

[golddiggie]

I don't think that your actual question was all that complicated (or too complicated to understand)... In that case, on a single vSwitch you would want to use the VLAN tagging, making sure it matches the tags on your physical switches, where the physical cables are connected, to route the traffic from VM port groups as needed... It will be a more complicated configuration, for certain, but still within the realm of 'doable'... I would suggest using enough pNIC's for that vSwitch to ensure good traffic flow. It also means that whoever is in charge of the LAN/network side of the environment will know to not make changes to those switch ports without notifying you far enough ahead of time... Nothing worse than to have the Network admin (or that group) make changes to switches, without letting you know, and taking x VM's off the network because of it... A situation like that would also do well to have redundancy just in case you have a switch fail on you, or have to go down for maintenance/updates.

VMware VCP4

Consider awarding points for "helpful" and/or "correct" answers.