Wired Intelligent Edge

 View Only
last person joined: 15 hours ago 

Bring performance and reliability to your network with the HPE Aruba Networking Core, Aggregation, and Access layer switches. Discuss the latest features and functionality of your switching devices, and find ways to improve security across your network to bring together a mobile-first solution
Expand all | Collapse all

AOS-CX VSF UPLINK

This thread has been viewed 46 times
  • 1.  AOS-CX VSF UPLINK

    Posted Aug 22, 2022 02:08 PM
      |   view attached
    Hi,

    I am creating a 6 member AOS-CX VSF stack with 6300's in a ring topology. The devices connected to the 6300 stack will be end user workstations and AP's. The 6300 stack will be configured to have a master and standby nodes with the remaining 4 nodes being members. The 6300 stack will have uplinks to my 5406R 2 member VSF stack (AOS).

    Question
    What is the best practice in cabling physical uplinks from the 6300 VSF stack to the 5406 VSF stack? Do I just need uplinks from the master/secondary nodes or all nodes? I have the capacity for as many 10GB SFP+ as necessary. 

    Thanks


  • 2.  RE: AOS-CX VSF UPLINK

    Posted Aug 23, 2022 01:53 AM
    A ring topology is the recommend according to this whitepaper:

    https://community.arubanetworks.com/HigherLogic/System/DownloadDocumentFile.ashx?DocumentFileKey=ed435687-5634-4891-b857-de9eec710512



  • 3.  RE: AOS-CX VSF UPLINK

    Posted Aug 23, 2022 09:26 AM
    Yes. I posted that in the first sentence. You are likely making the comment based on my Visio abilities :)


  • 4.  RE: AOS-CX VSF UPLINK
    Best Answer

    EMPLOYEE
    Posted Aug 23, 2022 03:09 AM
    Hello

    The VSF Configuration Guide has the following statement on page 10.

    https://www.arubanetworks.com/techdocs/AOS-CX/10.10/PDF/vsf.pdf

    "In a standard deployment, uplinks should be from primary and secondary"

    It is good to keep in mind that only the designated primary and secondary can be VSF conductors. If both of them fail at the same time, the remaining member switches will reboot, wait for conductor and keep their network ports down. They will not elect new conductor among them and operate as stack.

    Also if there is a stack split and primary and secondary are in one fragment, while in the other fragment there are only member switches, the fragment with the member switches will reboot and wait for a conductor with disabled ports. Again no new conductor will be elected among them.

    If you think that the uplink bandwidth of the conductor and secondary is not sufficient or if you want to avoid east west traffic within the stack (first send from member to conductor and then from conductor to 5400R) you can add uplinks from every member as well. But the main uplinks (with redundancy, for example LAG) should be from conductor and secondary because this will ensure that the stack remains operational in case of a failure.


  • 5.  RE: AOS-CX VSF UPLINK

    Posted Aug 23, 2022 09:55 AM
    Thank you @Emil_G.

    If I use 2x10GB SFP+ uplinks from primary and secondary ( 4 x 10GB SFP+) nodes, would only the links from the primary be active (20GB) or would all 4 links be active (40GB) if I create a single LAG from the 6300?

    Thanks


  • 6.  RE: AOS-CX VSF UPLINK

    EMPLOYEE
    Posted Aug 23, 2022 10:11 AM
    Hello, 
    If you create a LAG consisting of 4x10Gbps ports (2 from primary and 2 from secondary) all the 4 ports will be forwarding data traffic. The standby role of the secondary relates only the management and control plane. The data plane is fully distributed and active on all members.
    You may see some unequal distribution of traffic across the ports due to the LAG hashing algorithm but definitely all the ports will be active.


  • 7.  RE: AOS-CX VSF UPLINK

    Posted Aug 23, 2022 10:15 AM
    Awesome. Thanks!


  • 8.  RE: AOS-CX VSF UPLINK

    Posted Aug 26, 2022 01:29 PM
    Hey @Emil_G,


    In order to reduce east-west traffic, would a solution like this work for uplinks? I think multiple uplinks from the members is overkill for me. Wasn't sure if the member uplinks need a second LAG or if they should all be part of LAG1

    Master - 2 x 10GB - LAG1. Each connection to each core switch
    Standby - 2 x 10GB - LAG1- Each connection to each core switch
    Member 3 - 1 x 10GB - LAG2 - Connection to core#1
    Member 4 - 1 x 10GB - LAG2 - Connection to core#2
    Member 5 - 1 x 10GB - LAG2 - Connection to core#1
    Member 6​ - 1 x 10GB - LAG2 - Connection to core#2


  • 9.  RE: AOS-CX VSF UPLINK

    EMPLOYEE
    Posted Aug 27, 2022 08:32 AM

    Hello
    Since Core1 and Core2 are members of a VSF stack and all 6 6300 are also a VSF stack all uplinks could be in the same LAG. A LAG can have up to 16 physical ports on CX 6300 and 8 on the 5406R (assuming latest firmware).
    Having 2 LAGs between both VSF stacks will introduce problems with STP. If STP is enabled then one of the LAGs will be blocked. If it is not enabled you will have a loop. It is possible to tag different VLANs on both LAGs and configure different MSTP instances.
    However this is not a typical scenario with switch stacks since one of the main benefits of stacking is to avoid complicated MSTP configuration.

    Having LAG member port from every member should eliminate east-west traffic. The VSF stack has local-first forwarding for unicast traffic forwarded  via LAGs. That means if a member switch has a local port member of the LAG the traffic is forwarded to it and the LAG hash algorithm is ignored.