Route based on physical NIC load & failback

Ran into an interesting little problem involving routebased on physical NIC load & failback. With all the other NIC teaming method, if failback is set to no, it'll ignore the nic when it comes back online until the VM is rebooted. Physical NIC load is different. It's dynamic. It shouldn't care what failback is set to (I think). Anyone ever seen any documentation stating one way or other involving physical NIC load & failback?

Find more posts tagged with

Free for TechExams community: Cybersecurity salary guide

Compare cert salaries and plan your next career move

Button

Comments

jibbajabba

When you set failback to "yes" - did you specify a failback order ?
As far as I understand - if you set the order to nic0/nic1/nic2, and nic0's load is to high, it should fail back to nic1 once it is backup ...

dave330i

Here's my theory on this:

I believe failback should be "no". Physical nic load checks the load on the nic every 30 sec, so when a failed nic comes back online, it should wait 30 sec, check the load on the nics and balance the load across the nics accordingly. By introducing a 30 sec wait buffer, it can help mitigate network flapping.

The wrinkle in this is what does physical nic load algorithm do when 1 nic fails and it's down to 1 nic? Does it keep checking the nic loads every 30 sec?

Too bad I don't have a lab I could test it against. It's a simple enough test.

meadIT

One additional scenario to keep in mind on the failback being set to "yes." If your Network Failure Detection is set to Link State only, there are scenarios where the link state will come back up before the port is ready to transmit traffic, which could cause a period of inaccessibility.

jibbajabba

meadIT wrote: »

One additional scenario to keep in mind on the failback being set to "yes." If your Network Failure Detection is set to Link State only, there are scenarios where the link state will come back up before the port is ready to transmit traffic, which could cause a period of inaccessibility.

Good point, especially without portfast.

dave330i

@meadIT,

Would you consider this a VCDX defense type question or too easy of a problem?

meadIT

Yeah, I could see this question come up during a VCDX defense. In your design, if you choose one or the other, I would spell out your reasoning for choosing that setting. Anything that shows your thought process will help you score points.

Quick Links

All Categories
Recent Posts
Activity
Unanswered
Groups
Best Of