Routing

last person joined: 3 days ago 

Ask questions and share experiences about ACX Series, CTP Series, MX Series, PTX Series, SSR Series, JRR Series, and all things routing, including portfolios and protocols.

  • 1.  MC-LAG back to back

    Posted 09-08-2014 05:24
      |   view attached

    Hi, 

     

    Can you please confirm the doability of having a back to back MC-LAG. 

    See the the attached drawing.

     

    Regards

    Attachment(s)

    pdf
    MC-LAG.pdf   17 KB 1 version


  • 2.  RE: MC-LAG back to back
    Best Answer

    Posted 09-11-2014 23:16

    I did it and it works fine, It is supported.

     



  • 3.  RE: MC-LAG back to back

    Posted 01-29-2016 11:17
      |   view attached

    I have 2 pairs of QFX5100s that I have setup for back to back MCLAGs and on the surface it works, except for issues running spanning-tree.

     

    It seems that the QFXs do not send BPDUs out all members of ae/mc-lag interfaces.  I have a similar setup to the one listed in this form except that I have crossed linked the QFXs.

     

    For example in my setup I want Building1_Switch _1 to be root for both RSTP and VSTP, however Building1_switch_2 also thinks it is root (due to no spanning-tree running on mc-lag peer as per best practice).

    The interesting piece is Building2_Switch_1 sees the proper root bridge via AE3, however Building2_Switch_2 sees Building1_switch_2 as root.

     

    I'm guessing it is the way the QFXs hash BPDU out the AE members.

     

    Anyone run across this before, anyone know of a fix, is this an unsupported model in the Juniper mc-lag world???



  • 4.  RE: MC-LAG back to back

    Posted 01-30-2016 07:03

    Hello,

     

    @Doug Johnston wrote:

     

     

     

    I'm guessing it is the way the QFXs hash BPDU out the AE members.

     

    Anyone run across this before, anyone know of a fix, is this an unsupported model in the Juniper mc-lag world???

    Only LACP and uBFD is sent out of ALL LAG ports. Everything else is sent out via single LAG member interface only.

    AFAIK, this is not specific to JNPR and other vendors follow this method too.

    You don't want Your switch to do half a dozen or a dozen STP changes when it receives 6-12 identical BPDUs, each with TCN set. Since STP protocol itself does not have built-in mechanisms/is not intelligent enough to do it only once, other preventative measures need to be taken.

    Locally-originated traffic is NOT hashed, a single LAG member interface is chosen instead.

    Normally this is lowest numbered port.

    Try to rebuild Your LAG such that lowest numbered port is active in A/P MCLAG, then re-test and report back.

    HTH

    Thx

    Alex

     



  • 5.  RE: MC-LAG back to back

    Posted 09-28-2016 07:36

    Hi Doug, 

     

    I am trying to build a back to back mc-lag for testing purposes.  I wonder if  you could share the mc-lag  related configuration in your set up.   when we configure  mc-lag only on a pair of devices   towards one device  , the two links on that other device  are configured as a  lag interface.   From the topology set that you provided , I do not get  a clear  view how this is configured. 

     

    Thanks,