Wired Intelligent Edge

 View Only
  • 1.  VSX Downstream LAG

    Posted Nov 05, 2021 12:34 PM
    Hello,

    I am currently working on setting up VSX on two pairs of 8325s. 

    I have VSX configured on them. These two clusters will be connected to each other, but one will also be connected to a 6300 Series and the other will be connected to a different 6300 Series switch in another location. 

    My question is, will I need to configure the downstream LAG on the interfaces that connect the two VSX Clusters and the separate 6300 each 8325 is connected to as well?


  • 2.  RE: VSX Downstream LAG

    Posted Nov 05, 2021 01:24 PM
    Edited by Ivan_B Nov 05, 2021 06:01 PM
    Hi!

    It would be a great help if you could illustrate your question with a network topology diagram. 






  • 3.  RE: VSX Downstream LAG

    Posted Nov 05, 2021 02:46 PM
    Edited by sfpit1202 Nov 05, 2021 05:11 PM
      |   view attached
    This is a diagram of what I was trying to do...

    I have 2 J479A and 2 J 581A. I have 1 of each that I am setting up as a VSX Cluster. Right now, I have them configured with VSX. I am trying to figure out which connections need to be configured for a downstream LAG.  The purple links are the VSX links. The red link with .1 and .2 is the link connecting the two VSX Clusters and . 9 on the top goes to a 6300 Series switch and .3 on the bottom goes to a different one. 

    Do I need to configure the interfaces on the top (.1 and .9) and the interfaces on the bottom (.2 and .3) for the downstream LAG? I guess I am just a little confused, but if I am understanding right, the LAG links are/ can be the uplinks to other switches, but is that what I would be doing to the link between the two clusters even though they are both VSX Clusters? I am just going to create another LAG interface and assign each port to their LAG interface and the port also has an mtu setting .





  • 4.  RE: VSX Downstream LAG

    Posted Nov 05, 2021 05:18 PM
    Hi, I'm not able to understand what you wrote and, at least from what I am capable of seeing, also the attached picture is not very useful to umderstand what you want to achieve.

    In any case, a VSX LAG (Multi-Chassis LAG) is defined and used when you need to connect a VSX Cluster (a pair of Aruba 8325, as example) to an upstream/downstream peer device (clearly IF you accept the assumption that that peer needs to be concurrently connected to both VSX Members possibly via a LACP LAG). It's pretty clear that IF the peer device is a VSX Cluster too, on its ends you will need a VSX LAG too (the line of reasoning is: the VSX connects to its peers possibly through VSX LAGs). If the peer device is a standalone switch...it means that you'll simply need to configure a classic LAG (with LACP).





  • 5.  RE: VSX Downstream LAG

    Posted Nov 05, 2021 06:41 PM
      |   view attached
    I am trying to understand where I need to configure the VSX LAG Links. This is a better image of my diagram I hope. 

    This is on my backbone. I am going to be using two VSX Clusters (Each cluster has a J479A and a J518A). They will have a connection between the two VSX Clusters and then VSX Cluster 1 will be connected to a single 6300 Series Switch (JL661A) and VSX Cluster 2 will be connected to a stack of 6300 Series Switches (JL661A, I am going to use VSF to stack these). I am going to be using OSPF and I am going to configure each device as a DHCP-Server. 

    I think I get it. The link between the two VSX Clusters will be a VSX LAG Link. The link from each VSX Cluster to the 6300s doesn't have to a LAG Link.





  • 6.  RE: VSX Downstream LAG

    Posted Nov 06, 2021 09:13 AM
    Edited by parnassus Nov 06, 2021 09:16 AM
    Hi!

    "I think I get it. The link between the two VSX Clusters will be a VSX LAG Link. The link from each VSX Cluster to the 6300s doesn't have to a LAG Link."

    Exactly, better rephrased:

    "I think I get it. The link between the two VSX Clusters will be a VSX LAG Link. The link from each VSX Cluster to the 6300s doesn't have to be necessarily a VSX LAG Link."

    but the fact that, from a VSX Cluster, you are connecting a downstream/upstream peer device...it implies that you want the peer device to be connected to both VSX Cluster's members (Primary and Secondary) to benefit of the enhanced resiliency that a similar link brings to the topology you're building...that's to say that connecting a peer switch to a VSX Cluster using just a single upstream link (to Primary or to Secondary) is a bad design WHEN you can connect the very same peer switch to both VSX Cluster's members concurrently (using VSX LAG on the VSX side and a normal LAG on the peer switch side).

    Clearly all the above is generically valid IF you're not creating de-facto a topology loop (Switch "Z" connected to VSX Cluster "A" connected to VSX Cluster "B" connected to Switch "Z" <- Loop) and, also, the same is true if you, at one end, have a VSF Cluster instead of a VSX Cluster...or two VSF Clusters.

    ------------------------------
    Davide Poletto
    ------------------------------