Hoping to clear a couple things up.
1. Say I have a vswitch with two active active pnic adapters. On this vswitch I have two IPs assigned, one portgroup for the service console, the other for vmkernel. (they both route to the same gateway)
Is is an acceptable configuration to keep an active / active configuration for both port groups. If so why? It appears to work just fine, but I've been challenged on this scenario.
Or should I have something like this:
SC - vmnic 0 active, vmnic 1 standby
VMkernel - vmnic 1 active, vmnic 0 standy.
We are using the default vswitch properties, so
2. Say I have the same scenario on a different vswitch that is just for virtual machine traffic. Two active / active pnics.
Is traffic always going down both pnics? If yes, does traffic for one vm always stay on the same pnic?
When i look at esxtop I just see one pnic assigned to the vms.
Using vsphere and active/active pnics are spread across two switches.
Thanks!
In regards to question 2, the default policy on the standard vSwitch will result in assigning the virtual nics to a different port alternating between switches.
So VM 1 will get assigned a virtual port on pnic 1, VM2 will get assigned a port on pnic 2, VM3 to pnic1, and so on. The VM will continue to use the same pnic until it moves off the host, and possibly a power off and on might change it, I would have to look that up to be sure.
In regards to question 1, your config will work just fine, but have you looked to see how traffic is being distributed across both nics (or not) in esxtop for example?
Is the vmkernel for vmotion? Storage? Its generally considered a better practice to separate these on different nics in an active.standy config to better balance the load. There are also security concerns to consider but it sounds like that is not a concern for you if your storage or vmotion traffic is routable.
Assuming your vmkernel connection is for vmotion, what many people do is have vmnic0 be sc primary and vmotion failover/secondary and vmnic1 primary for vmotion and sc failover/secondary.
Google Ken Cline who has some lengthy blog essays that go into great detail on the considerations and concerns for this and related nic configurations for these kinds of questions.
1. Say I have a vswitch with two active active pnic adapters. On this vswitch I have two IPs assigned, one portgroup for the service console, the other for vmkernel. (they both route to the same gateway)
Is is an acceptable configuration to keep an active / active configuration for both port groups. If so why? It appears to work just fine, but I've been challenged on this scenario.
Or should I have something like this:
SC - vmnic 0 active, vmnic 1 standby
VMkernel - vmnic 1 active, vmnic 0 standy.
We are using the default vswitch properties, so <span class="Interface"><span class="Interface">
The distribution of the network traffic with a single Service Console port and a single VMkernel port on a vSwitch with 2 active uplinks and the default load balancing setting will result in the same spread of traffic as the active/standby config you described. All of the Service Console traffic will use a single uplink and allof the VMkernel traffic will use the other uplink. If one of the uplinks fails, then both the Service Console and the VMkernel would use the remaining uplink until the other is restored.
With the configuration you shared, you decide which one of the uplinks is used for each, while the default load balancing setting will "pick" for you.
Net result, no performance difference advantage, just explicit active/standby order that must be set for each portgroup.
Dennis Bray VCI, VCP
