Frequent Disconnection by Devices

[NWNSM1]

We have users experiencing disconnections or roaming issues.  Attached is a sample from one user. 

APs are CW9166I, and clients are a mixture of devices (iPad/Apple TV/MacBook/HP Win11)

The environment is 2-story middle school with an AP in every classroom or office/media center.  Students have iPads, but again, this is not device or OS specific.  

The 29 APs are running two 5 Ghz radios, and many of the roams are between the two radios on the same AP.  

We experience a high number of unexplained disconnections, too. 

The environment is RF clean, and the Meraki dashboard indicates that channel utilization and interference are low. 

The APs are new, and we had been running MR42 or MR52 since 2018 without issues.  

[alemabrahao]

I advise you to conduct a site survey of your Wi-Fi network.

You can also check if 802.1r and client load balancing are enabled and disable them to see the result.

[PhilipDAth]

The first thing I check these days is whether the customer has an "RF Profile" configured.  It solves about 90% of the issues immediately by creating one.

https://documentation.meraki.com/Wireless/Operate_and_Maintain/User_Guides/Radio_Settings/RF_Profile...

I mostly use the "Open Office Profile", but in your case the "Classromm Profile" might be best.

Make sure you make the profile the "default", or apply it to all your APs.

And while you are there, make sure you turn on AI channel planning and "busy hour".

https://documentation.meraki.com/Wireless/Operate_and_Maintain/User_Guides/Radio_Settings/AI_-_RRM

[NWNSM1 · Accepted solution]

I did have an RF Profile enabled and applied.  I opted to ask ChatGPT to fix my network, and by looking at a few specific clients and as a whole, it made suggestions.  It was able to suggest changes that appear to be working. Below is the ChatGPT summary. 

Multiple clients across the SCHOOL SSID experienced intermittent disconnects and excessive roaming in a high-density classroom environment using Meraki CW9611 APs. Analysis of Meraki event logs showed frequent same-AP roaming and rapid disassociate/reassociate cycles despite strong RSSI/SNR, ruling out RF coverage, interference, DFS, or client-specific issues. Root cause was determined to be aggressive roaming behavior related to dual 5 GHz operation combined with 802.11r (fast roaming) in a low-mobility environment. Mitigations applied included disabling 802.11r on SCHOOL, maintaining 20 MHz channels, controlled TX power (11–17 dBm), manual RX-SOP, and disabling client load balancing. Post-change validation shows stable client associations (Windows and iOS), elimination of same-AP roaming storms, and behavior consistent with normal client operation. Issue considered resolved; no further RF changes recommended.

[AJAY2]

So do you say to disable 802.11r if Dual 5Ghz radio is enabled in the APs ?

[NWNSM1]

I am not suggesting that anything that worked for me will be a solution for someone else.  What I recommend is to take your logs and settings from Meraki and paste them into ChatGPT: small changes, new logs, good natural language prompts and then rinse and repeat.  The only tools I used were the Meraki dashboard, ChatGPT, and WiFi Explorer. 

ChatGPT didn't require any fancy parsing the log files. I used event viewer, AP logs, device logs, I cut and pasted HTML pages and screenshots when they was the best info I had and it never once had a problem deciphering anything. Again, I asked questions and tried to do things in a logical order but could change streams on a dime with a quick prompt letting ChatGPT we were looking at a new issue, device, or building. I even cut and pasted end user tickets that did a sometimes hilarious description of the experience.

ChatGPT did things like identify device types from MAC addresses it found in the logs and provide info on their known behavior and design solutions accordingly. ChatGPT knows not just good wireless design but Meraki wireless (and networking) and it even the different behaviors of the exact AP model I had deployed. 

What I did was condense hours/days/weeks of searching the web and trying to piece together solutions. 

and essentially fast-track me to the end.  I'm one month out now and there have been no network tickets or problems needing resolved. 

I am now using ChatGPT to build an enterprise level NOC of my Meraki network with using the API and Python. 

(I should note that any sensitive data was stripped away from chatGPT prompts.) 

[AJAY2]

Thank you & noted.

I am now using ChatGPT to build an enterprise level NOC of my Meraki network with using the API and Python.  Great! So are you using any Open-source software, I am too looking to build a basic sample NOC solution.

[NWNSM1]

Python has Meraki libraries which I installed and ChatGPT wrote everything in Python.  Currently it's raw code but when I ready to put it into production, it will package it up and secure the API keys and such.  (Side note: I know zero about Python and ChatGPT is doing all of the lifting.)

This was my first prompt and the response, followed by some refinements, all took place over lunch at my desk:

Do you know how to work with the Meraki API?

Yes — very familiar with the Meraki Dashboard API 👍

Both:

🌐 v1 REST API

🐍 Python SDK (meraki library)

🔐 API key authentication

📊 Org / Network / Device endpoints

📡 Wireless analytics

📈 Client usage

🔄 Config automation

🚨 Alerts & monitoring

🧪 Action batches

If you’re running Meraki Wi-Fi in a 1:1 iPad district like you described earlier, the API can be extremely useful for:

Finding high-utilization APs

Pulling client counts per AP

Checking channel utilization

Detecting rogue devices

Exporting usage stats

Automating SSID changes

Pulling WAN utilization from MX

Monitoring packet loss / latency

Identifying top talkers