Route based on physical NIC load & failback
dave330i
Member Posts: 2,091 ■■■■■■■■■■
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?
2018 Certification Goals: Maybe VMware Sales Cert
"Simplify, then add lightness" -Colin Chapman
"Simplify, then add lightness" -Colin Chapman
Comments
-
jibbajabba Member Posts: 4,317 ■■■■■■■■□□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 ...My own knowledge base made public: http://open902.com -
dave330i Member Posts: 2,091 ■■■■■■■■■■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.2018 Certification Goals: Maybe VMware Sales Cert
"Simplify, then add lightness" -Colin Chapman -
meadIT Member Posts: 581 ■■■■□□□□□□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.CERTS: VCDX #110 / VCAP-DCA #500 (v5 & 4) / VCAP-DCD #10(v5 & 4) / VCP 5 & 4 / EMCISA / MCSE 2003 / MCTS: Vista / CCNA / CCENT / Security+ / Network+ / Project+ / CIW Database Design Specialist, Professional, Associate
-
jibbajabba Member Posts: 4,317 ■■■■■■■■□□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.My own knowledge base made public: http://open902.com -
dave330i Member Posts: 2,091 ■■■■■■■■■■2018 Certification Goals: Maybe VMware Sales Cert
"Simplify, then add lightness" -Colin Chapman -
meadIT Member Posts: 581 ■■■■□□□□□□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.CERTS: VCDX #110 / VCAP-DCA #500 (v5 & 4) / VCAP-DCD #10(v5 & 4) / VCP 5 & 4 / EMCISA / MCSE 2003 / MCTS: Vista / CCNA / CCENT / Security+ / Network+ / Project+ / CIW Database Design Specialist, Professional, Associate