Wireless Access

 View Only
  • 1.  Cluster and LMS/BKP-LMS IP

    Posted May 23, 2019 11:06 AM

    I am designing a network with two 7220 controllers in one cluster. both controllers are in the same vlan and I configured VRRP between them. I though it would be a good idea to configure LMS-IP to be the virtual VRRP IP address and leave BKP-LMS IP empty, but I found in Aruba documentation they are using controller physical IP address, and not VRRP, for LMS IP. I beleive this is due to the fact that AP will receive node list eventually and will point to this node list, and not LMS IP, but LMS-IP redundancy is needed also.

     

    in my scenario of two controllers only, which option is recommended:

     

    1- LMS IP: VRRP

    BKP LMS IP: Empty

     

    2- LMS IP: Controller 1 IP

    BKP LMS IP: Controller 2 IP

     

    3- LMS IP: VRRP

    BKP LMS IP: Controller 1 or Controller 2

    (in case of IP conflict with VRRP IP address)

     

    Thank you

     

     

     



  • 2.  RE: Cluster and LMS/BKP-LMS IP
    Best Answer

    Posted May 23, 2019 04:57 PM

    4.  LMS-IP Blank.

     

    The APs only have to discover the controller VRRP and the nodelist will be distributed to them.  Your "cold boot" discovery method (dns or dhcp option) should be pointing to the VRRP between controllers.  When the AP hits the cluster, the nodelist will be pushed to flash on the AP and will survive a reboot.  The dns or dhcp option is really only needed for the first time an AP is provisioned and the lms-ip is only if you want an AP to find a cluster besides the first one that it finds on initial cold boot.  backup-lms ip would be for the paranoid who need a second redundancy option beyond  what a cluster provides.



  • 3.  RE: Cluster and LMS/BKP-LMS IP

    Posted Feb 24, 2020 01:12 PM

    We have 2 clusters one in each data center wanted to have data center redundancy. We configured the primary LMS for the primary VRRP cluster IP and the backup IP to the VRRP IP for the other cluster / data center. While we didn't test to see if the AP's would use the backup LMS what we did see is some AP's ended up over on the other cluster even though the Provisioning IP in the provisioning page is the VRRP IP for the primary cluster. I tried to provision them again they are stuck over at the other data centers cluster I assume ones the node list is saved in the flash of the AP can't change it unless we factory reset. I suspect those AP's ended up over there during the migration process we had the primary / back up LMS ip at the group level AP system profile already setup. That would be the only time the AP would even look at the LMS. When using  a cluster LMS and backup LMS is not even used no reason to even configure it?



  • 4.  RE: Cluster and LMS/BKP-LMS IP

    Posted Feb 24, 2020 01:24 PM

    You should not configure a backup LMS when you have a cluster that already provides redundancy.  Why? because as you just described, it makes troubleshooting too complex, but for what?  Controllers don't fail often, but if you had to replace one, you could remove it from the cluster and the APs would rebalance.

     

    If  all of your controllers failed, would you want your access points to connect to a different datacenter?  What if you accidentally disrupted the network upstream of some of your access points and only half of them appeared at the datacenter and users could not roam between those access point and access points that still had connectivity to your primary cluster?  Again, a cluster IS your redundancy.  Backup LMS with a cluster creates too many opportunities for "why are my access points on that cluster".  Keep it simple.



  • 5.  RE: Cluster and LMS/BKP-LMS IP

    Posted Feb 25, 2020 12:17 AM

    Reading from the guide this should work and if Preempt is used it should go back to the primary LMS. From page 335 of 8.3.0.0 user guide. Were on 8.3.0.10 These AP's the 34 I can't get them to go back even with power cycle. They can find the cluster node IP and it works stuck on that cluster. Is this normal behavior for cluster once an AP is tied to a cluster can't be moved to another one? 

     

    AP Failover to Different Cluster
    Starting from ArubaOS 8.0.0.0, an AP can fail over between clusters. Redundancy across geographically
    separated data centers are supported. An AP terminates on an AAC in a cluster. If a member in the cluster fails,
    the AP will fails over to the S-AAC in the same cluster. If the AP is unable to establish communication with any
    of the members in the first cluster, then it terminates on another cluster setup in the backup data center. It
    terminates on another cluster only if the other cluster member IP is provided in the AP system profile as
    backup LMS.
    For example, a cluster with four managed devices is deployed in the West Coast data center. Similarly, a cluster
    with four managed devices is deployed in the East Coast data center. An AP is configured to have a primary
    termination on the West Coast data center and backup termination on the East Coast data center. If a
    managed device fails in the West Coast data center, then the AAC moves to another managed device in the
    same data center. However, if the entire West Coast data center is inaccessible to the AP, then it fails over to
    the East Coast data center.