vSphere Standard Switch - How does failover work

[ToTheCloud]

Hello,

I have kind of a theory question for you VMware guru's.  I have four ports on my ESXi host and I would like to make sure that in the event of a link failure, the network traffic gets moved over to another link.  Currently, I have all four adapters marked as Active to load balance traffic among the four ports.  My question is, what is the point of having a standby adapter?  In the event of a link failure, does an active switchport not failover to another active port?  If an active port can failover to another active port, why would you ever want to set an interface into standby mode, rather than utilize it to loadbalance traffic?

Thanks for any insight in advance!

-ToTheCloud

[vmroyale]

Note: Discussion successfully moved from VMware vSphere Hypervisor to VMware vSphere™ vNetwork

Standby adapters can be useful where you want a single nic to have a dedicated function. For example, a vSwitch with two physical NICs and two VMkernel port groups - one used exclusively for FT and one used exclusively for vMotion. But in the event of a single NIC failure here, the standby adapter can be used by both port groups.

[ToTheCloud]

Ok, so the standby adapter would be useful if I only had 3 ports and two vSphere Standard Switches.  At that point, when a port to either of those switches would fail, I would be able to have either one of those switches fail over to the standby adapter (since you cannot have an active adapter on two switches at the same time)?

Also, just to clarify, if I do have all four adapters active on the same vSwitch, if one link goes down, traffic would failover to another active adapter if I didn't have a standby adapter specified, correct?

Thanks!

-TotheCloud

[vmroyale · Accepted solution]

Ok, so the standby adapter would be useful if I only had 3 ports and two  vSphere Standard Switches.  At that point, when a port to either of  those switches would fail, I would be able to have either one of those  switches fail over to the standby adapter (since you cannot have an  active adapter on two switches at the same time)?

Correct - if you had two vSwitches, each with one NIC. The standby adapter could only be assigned to one vSwitch at a time.

Also, just to clarify, if I do have all four adapters active on the same  vSwitch, if one link goes down, traffic would failover to another  active adapter if I didn't have a standby adapter specified, correct?

Yes.

Message was edited by: vmroyale to update based on a mistake (which is now striked-through)

[ToTheCloud]

Excellent!  Thank you for explaining this!

[ToTheCloud]

@vmroyale, sorry to bump this back up, but I just tried this in my lab environment, and I cannot seem to share the same port as a standby one.

For example, I have 3 ports.

1 active port to vSwitch0

1 active port to vSwitch1

Now I would like to set my 3rd port to be in standby for either vSwitch0 or vSwitch1 to provide redundancy to either of those in the event that one or the other went down.  Once I add the 3rd adapter as a standby adapter to vSwitch0, I am unable to select it as a standby adapter for vSwitch1 (I receive, "One or more of the selected adapters are attached to other vSphere standard switches.  Are you sure you want to remove them from these vSphere standard switches and dd them to this one?"

I guess I still don't understand what is going on here

[vmroyale]

My apologies - I read your text wrong above. I read "when a port to either of  those switches would fail, I would be able to have either one of those port groups fail over to the standby adapter" which went back to my earlier example with vMotion and FT networking.

Your testing shows the expected behavior.

[ToTheCloud]

So using a standby port would really only be beneificial when using port groups with a vSphere Distributed Switch?  I guess I don't understand why someone wouldn't make them active all the time (to take advantage of load balancing instead of just having the port sit until a failure occurs), unless they were low on ports and would want to provide some redudancy for a setup like in my example above (which doesn't seem to be possible with vSphere standard switches).

-ToTheCloud

[vmroyale]

Take the FT and vMotion example again. I would want these separated, so that FT can do its thing on a dedicated NIC and vMotion can do its thing on another dedicated NIC. Now I don't have to worry about vMotion impacting my FT interval, and I don't have to worry about FT impacting vMotions. This is a great setup, when things are working great with both NICs. In a situation where I lose a NIC, I would rather have vMotion and FT both work, vs possibly losing one feature entirely.

For virtual machine networking, I don't know if the standby adapter makes much sense. Like you said, just use them all and don't worry about it. But with the VMkernel networking it can be useful in certain situations.

[ToTheCloud]

Brian Atkinson wrote:

Take the FT and vMotion example again. I would want these separated, so that FT can do its thing on a dedicated NIC and vMotion can do its thing on another dedicated NIC. Now I don't have to worry about vMotion impacting my FT interval, and I don't have to worry about FT impacting vMotions. This is a great setup, when things are working great with both NICs. In a situation where I lose a NIC, I would rather have vMotion and FT both work, vs possibly losing one feature entirely.

So in that case, how do you assign the same (or third) NIC to be a standby adapter to failover either the FT or vMotion switch (whichever failed first?  From what I understand from your response, I should be able to create two seperate vSwitch's, and assign the 3rd NIC to both vSwitch's as a standby adapter.  In the event that one vSwitch fails, the standby adapter would kick in.  In this case, if both vSwitch's failed, whichever vSwitch failed first would use the standby adapter and then the second vSwitch to fail, would stay failed since I wouldn't have another adapter on standby (but in that case, if I had 4 NICs, I would just assign 2 Active adapters to each vSwitch and just do away with the Standby adapter).

Sorry to be slow on understanding this...

[vmroyale]

I think it might just be a terminology issue between port groups and vSwitches. The setup I have previously described is a single vSwitch with two portgroups created in it. This vSwitch would also have two active Network Adapters assigned to it. The vSwitch propeties would show the following port groups:

One port group is used for vMotion and is set up to use a dedicated active adapter (NIC) and the other is used for FT and is also set up to use a dedicated active adapter. In this setup, the FT port group would look like this:

The "Overriide Switch Failover Order" option is used above to allow a single physical NIC to be dedicated to this port group.

The vMotion port group looks like this, with the active and standby swapped from the previous screenshot:

Using the Failover Order options allows the separation, but it per-vSwitch with the standby NICs. This setup allows the two NICs to provide failover for each other.

You can't have the same standby adapter assigned to two separate vSwitches. If you think about how your networks are split up and isolated (hopefully) this common standby adapter would be an unlikely option.

[ToTheCloud]

Thank you sooo much for this reply vmroyale!   I really appreciate the time you put into the replay, it really helped clarify the use of this!