APs dropping from VSZ on firmware 6.1

[Luis]

We are using a VSZ, we noticed that when the controller reaches 100% of CPU, ALL APs drop (about 60 APs), then start reconnecting slowly (and keep dropping from time to time). We didn't have any issues on firmware version 5, and no changes in the network has been made, but once we upgraded to 6.1, this started happening. We have drastically increased the resources in the VM, they are more than enough and more than the minimum requirement.

We are really struggling with this.

Thanks

[Luis]

Another piece of information.

Since this is a school, I've been taking notes on the more massive disconnections.

They always happen 5 min after the end of a class. Seems like a massive client roaming is triggering the CPU to go 100% and cause the APs "dropping"

Are you aware of any different behaviour in this firmware version when massive client roaming occurs? The pattern is pretty clear now.

[Luis]

I guess the important question is...

With the firmware at 6.1 the CPU spiked to 100% during clients roaming between class times, causing a timeout in the communication between controller and APs. Since we downgraded back to 5.2.2, the CPU spikes to 67%. Maybe it was always behaving like this but since it didn't cause any issues, it went undetected.

The question now is. At the moment we have 115 APs, but we are planning to add an extra 100. How is this going to impact the CPU? in theory our setup should handle up to 500 APs.

Do you have any reason for the spikes during massive client roaming? is this normal or is it specific to us?

The only thing stopping us from adding the next batch of 100 APs is this issue.

Any insight would be appreciated.

[sanjay_kumar]

Hi @Luis 
Normally this issue should happen with just 115 APs and the number of clients you have. In 5.2.2 version, does the CPU always stays at 67%? Or it will be 67% (during the below scenario) and will be less during other times? If yes, then we can see the controller statistics and collect the logs to check the issue.

"They always happen 5 min after the end of a class. Seems like a massive client roaming is triggering the CPU to go 100% and cause the APs "dropping"

[Luis]

Hi Sanjay,

CPU at 67% only happens during class change, then it goes back down to 17%

Usually the CPU usage is low, but when most users roam at the same time, the CPU goes drastically up.

[sanjay_kumar]

Hi @Luis 
I believe CPI spike issue still remains the same on 5.2.2 as well but it is not striking to 100% which happened in 6.1.
We can still troubleshoot this in 5.2.2, please open a support case so that we can check the setup and collect the relevant logs to check the issue.