Wireless Access

Hello,

 

I would like to validate this kind of special deployment

 

I have a customer wth the following installation:

 

- 2 x VMMs running AOS 8.3

- 1 cluster of 2 x 7205 on HQ (same AOS version)

- 18 x 7008 controllers deployed on remote branches (same AOS version)

- 100+ APs

- DNS aruba-master pointing to the HQ cluster VIP

- On each branches the AP System Profiles are configured with the IP of the local controller as primary LMS and the VIP of the HQ cluster as backup LMS

 

During a recent upgrade (from 8.3.0.3 > 8.3.0.7) we had a strange behaviour of the APs when the branches 7008 were already upgraded but the HQ cluster was still running the old version, the APs were just flapping between the branches controllers and the HQ cluster and each time upgrading or downgrading the FW. When the HQ cluster was finally upgraded all APs register correctly to their own primary LMS.

 

- Is this kind of configuration really valid and supported ?

- What can be changed to mitigate the impact of temporary running mixed version ?

 

Kind regards

Christian Chautems

The branches should have dhcp options 43 and 60 pointing to the branch controller as an exception so that those access points at the branch do not connect to the HQ cluster.  If you point access points at a controller with a higher or lower firmware, the aps will upgrade or downgrade and reboot.  In your situation, they downgraded, and then when they were sent to their lms-ip, they were forced to upgrade, creating a loop.  You can use aruba-master to point devices at the HQ cluster, but use dhcp options 43 and 60 to override that at the remote sites, so they do not connect to the hq cluster.

Hello cjoseph,

 

Thank you for the quick answer that has confirmed what I was thinking. I will advise my customer to use the DHCP option for the remote branches.

 

Now outside the temporary mixed versions case is the settings of Primary LMS IP to a local MD and the Backup LMS IP to a cluster VIP valid when running same version on all controllers ?

 

What could be the restrictions of this settings ?

 

Kind regards

Christian Chautems

It is valid, from a configuration perspective.

 

Your users will definitely have reduced performance when the APs connect across the WAN to the HQ cluster, so you might want to make that failover something that is manual, instead of something that is automatic.  Controllers honestly do not fail often.  If there is a minor connectivity issue between the APs at the branch and the controller at the branch, you don't want the access points just failing over to the HQ cluster...  That will cause administrative confusion.

Thank you for your explanations.

 

Kind regards

 

Christian Chautems

I recommend removing the backup LMS when upgrading to avoid the flip flop between clusters



Thank you

Victor Fabian

Pardon typos sent from Mobile