Wireless Access

I have never run into this issue until now, but I have a customer that is seeing issues with Client Load Balancing.  I am hoping someone has seen this or could give some insight into what is going on.

Customer has Mobility Conductor (virtual) and 2 7220 controllers in an L2 cluster.  They have the defaults set for cluster redundancy and load balancing.  They have approximately 1k access points and, just under 12k clients.  They are running ArubaOS 8.7.1.2 on everything.

What we are seeing is that APs are load balanced well (about 500 on each controller) however, most clients are on one controller.  There are some clients on the other controller but the vast majority are on one (like 11500 on one and 70 on the other).  Expectation from the customer (and from myself) would be that some kind of load balancing would be happening.

Customer has been told that the load balancing won't happen unless the controller reaches the thresholds in the settings and, because they aren't there yet, most of the clients are on the one controller.  In the past, I have seen the load balancing working regardless of this so I checked my lab environment.  Unfortunately, my lab was way behind (8.6.0.2) in code so I couldn't compare directly.  In my lab, with 8 clients, I had 4 clients on each 7005 controller.  To me, that says that the balancing was working.  Additionally, I checked with another customer that had approximately 3500 clients over their 3 7240XM controllers and noticed that there were just over 1000 clients on each controller.  Again, this indicated to me that the client load balancing was working correctly.

So, as a test in my lab, I update my Mobility Conductor and controllers to 8.7.1.2.  After the upgrade, I observed that there were now 6 clients on one controller and 2 clients on the other.  Definitely not balanced.  As a further test, I upgraded to 8.7.1.3.  Same result.  Out of curiosity, I downgraded everything back to 8.6.0.2 and saw that there were still 6 clients on one controller and 2 on the other.

Now, I am sure that the load balancing may take some time, and I am going to look at it again in the morning to see where it is, but I am kind of at a loss.  One of the issues with this customer has always been when they get too many clients on one controller, they start seeing very poor performance.  Prior to the move from 6.5 to 8, we had manually load balanced everything between the controllers and we are starting to think we are going to have to move back to that again but wanted to check with the community before hand.

--
Jeremy R. Wirtz

I have never run into this issue until now, but I have a customer that is seeing issues with Client Load Balancing.  I am hoping someone has seen this or could give some insight into what is going on.

Customer has Mobility Conductor (virtual) and 2 7220 controllers in an L2 cluster.  They have the defaults set for cluster redundancy and load balancing.  They have approximately 1k access points and, just under 12k clients.  They are running ArubaOS 8.7.1.2 on everything.

What we are seeing is that APs are load balanced well (about 500 on each controller) however, most clients are on one controller.  There are some clients on the other controller but the vast majority are on one (like 11500 on one and 70 on the other).  Expectation from the customer (and from myself) would be that some kind of load balancing would be happening.

Customer has been told that the load balancing won't happen unless the controller reaches the thresholds in the settings and, because they aren't there yet, most of the clients are on the one controller.  In the past, I have seen the load balancing working regardless of this so I checked my lab environment.  Unfortunately, my lab was way behind (8.6.0.2) in code so I couldn't compare directly.  In my lab, with 8 clients, I had 4 clients on each 7005 controller.  To me, that says that the balancing was working.  Additionally, I checked with another customer that had approximately 3500 clients over their 3 7240XM controllers and noticed that there were just over 1000 clients on each controller.  Again, this indicated to me that the client load balancing was working correctly.

So, as a test in my lab, I update my Mobility Conductor and controllers to 8.7.1.2.  After the upgrade, I observed that there were now 6 clients on one controller and 2 clients on the other.  Definitely not balanced.  As a further test, I upgraded to 8.7.1.3.  Same result.  Out of curiosity, I downgraded everything back to 8.6.0.2 and saw that there were still 6 clients on one controller and 2 on the other.

Now, I am sure that the load balancing may take some time, and I am going to look at it again in the morning to see where it is, but I am kind of at a loss.  One of the issues with this customer has always been when they get too many clients on one controller, they start seeing very poor performance.  Prior to the move from 6.5 to 8, we had manually load balanced everything between the controllers and we are starting to think we are going to have to move back to that again but wanted to check with the community before hand.

--
Jeremy R. Wirtz