none
Dynamic(Switch Independent) NIC Teaming Mode Problem

    Question

  • Hi,

    We are using Switch Independent Dynamic distribution NIC Teaming configuration on Hyper-V 2012 R2 cluster. In the following graphs between 19:48 - 19:58 you can see clearly how dynamic NIC teaming mode badly affected the web servers response times.


    We have encountered this problem on the different cluster nodes that use different NIC chipsets Intel or Broadcom and configured with dynamic NIC teaming for VM Network connection.

    If we change the teaming mode as Hyper-V Port or remove NIC teaming for VM network as you can see from the graphic, response times are back to normal.

    Any ideas?

    Regards


    • Edited by HUNAL Tuesday, November 26, 2013 9:57 PM
    Tuesday, November 26, 2013 9:50 PM

All replies

  • Hyper-V Port is generally the recommended solution for Hyper-V installations.  Is there some reason why you do not want to use that?

    .:|:.:|:. tim

    Wednesday, November 27, 2013 5:20 PM
  • Hi Tim,

    Actually Microsoft recommends the new dynamic distribution mode in the "Windows Server 2012 R2 NIC Teaming (LBFO) Deployment and Management" guide http://www.microsoft.com/en-us/download/details.aspx?id=40319

    As well as Hyper-V Port mode doesn't provide bandwidth aggregation and NIC failure disrupts VM communication.

    Wednesday, November 27, 2013 6:02 PM
  • Hi,

    In the heavy inbound and outbound network load, If you are using the “Switch Dependent”  mode, you must configure your switch have the same teaming support mode.

    There have two scenario.

    1. “Switch Dependent” with LACP if your switch supports the aka 802.1ax.
    2. “Switch Dependent” mode with Static.


    Example of  Cisco® switch configuration when you using the “Switch Dependent” mode with “Static” mode:

    CiscoSwitch(config)# int port-channel1

    CiscoSwitch(config-if)# description NIC team for Windows Server 2012

    CiscoSwitch(config-if)# int gi0/23

    CiscoSwitch(config-if)# channel-group 1 mode on

    CiscoSwitch(config-if)# int gi0/24

    CiscoSwitch(config-if)# channel-group 1 mode on

    CiscoSwitch(config)# port-channel load-balance src-dst-ip

    Hope this helps.


    We are trying to better understand customer views on social support experience, so your participation in this interview project would be greatly appreciated if you have time.
    Thanks for helping make community forums a great place.



    Friday, November 29, 2013 10:39 AM
    Moderator
  • Hi Alex,

    Thank you for your answer, we would like to use switch independent mode so we can use separate switches, there is no heavy traffic as well.

    Actually we would like to find out the reason of that dynamic mode problem.

    Friday, November 29, 2013 11:30 AM
  • What else is running at that time? Is System Center Operations Manager in? You can use it to see what is flagged up.

    Is anti-virus running? Are you running intensive workloads on the shared storage at the same time?

    Are all nic drivers and firmware up to date? What about the switches, are they up to date?

    Sunday, December 01, 2013 10:23 AM
  • We have encountered the spikes when we live migrate the web and app. servers to the cluster nodes which are configured with the new dynamic NIC teaming.

    We don't use AV software and CSV latency between 6-9ms. All Intel and Broadcom NICs are updated to the latest firmware and drivers. When we remove teaming or change the teaming mode as Hyper-V Port on the same cluster nodes response times are back to normal.

    Sunday, December 01, 2013 2:16 PM
  • Do you have VMQ/RSS/SR-IOV enabled?
    Tuesday, December 03, 2013 10:08 PM
  • I am seeing the same thing where when our hardware load balancer runs a port 80 check is cycles between up and down.  Running wire shark we are seeing the Source MAC address change from the VM to the MAC of the Host.  If the Load Balancing mode is in either Dynamic or Address Hash modes the cycling continues.  If I change it to Hyper-V Port mode, it corrects the problem, but is not in the recommended configuration by the product team.    I have a case pending.  I you have any other info please share. 

    Rob McShinsky

    Monday, March 31, 2014 1:18 AM