Nope. Turns out there may be a bug in DAWN. I have made some further changes to work around this bug, and it appears to be working (albeit making DAWN really slow in provisioning any steering instructions).
(I have also discovered that, at least for my phones, getting DAWN to send frequent probe and beacon requests to these devices alone will make them hop to the desired APs all on their own under most cases. This along with DAWN's 'lethargic' steering instructions are enough for my needs)
You mentioned you have 4 APs, but do you have 4 routers? If not, that means some of these APs are located within the same router? If this is the case, then you may be getting hit with the same bug.
I raised a new issue in github. Doubt anyone will be dealing with it however judging from the number of open issues there that are getting ignored. Anyway the issue raised is located here: https://github.com/berlin-open-wireless-lab/DAWN/issues/243
1 Like
Thing is, I rolled back my tweaks (most of them anyway), and added some other tweaks, now that I have identified the root cause of my issue.
This issue may not apply to you. It only affects those with two or more APs (that DAWN is handling) located within the same router. e.g. a 2.4GHz and a 5GHz radio with the same SSID inside one router.
For a detailed description of the bug I have found see here: https://github.com/berlin-open-wireless-lab/DAWN/issues/243
If you are affected by this too, I can share my new tweaks. Mind you these tweaks will make DAWN really slow in giving out steering instructions to clients, but it works for my particular environment, and at least with these new tweaks DAWN won't issue inaccurate instructions that will actually sever connections (again only when you are hit with the bug I have identified).
my setup is 3 MX4300 set up as dumb AP, each broadcasting the same SSID on 2.4G and 5G. i've been back and forth on using the 3rd radio as mesh backhaul vs wires across hallways, but my roaming issues are independent of the backhaul method.
Based on my subsequent observations, i'm rolling back your changes to wireless.default_radio0.max_inactivity (back to 300) and dawn.@times[0].remove_probe.
Still awaiting better solutions for roaming assistance.