I deployed a mesh network with 10 MAPs and 2 RAPs which are 1522's (hazardous location APs) using a 4402 controller running 126.96.36.199M MESH.
First of all, I'm having problems with some of the MAPs taking sometimes a half an hour to join the mesh then drop out of the mesh and return with a "power loss" message. The power source of these MAPs is 120v power coming from a light pole which has a UPS backup. I'm assuming a small glitch in the power could cause this but I've been told by an onsite contact this is not likely.
Another issue I've been having is with the performance of the backhual interface. I have 2 RAPs with the same BGN but 6 MAPs use one RAP and none use the other. Some of the MAPs should see the childless RAP as a better path but for some reason don't use it. If I power cycle the RAP with children all of the MAPs migrate to the childless RAP. Is there a way to "load balance" RAPs without dividing the mesh using two BGNs?
Also, by default the backhaul power is set to 1 which is causing most MAPs to be the children of the single RAP. What is the rule of thumb for the number of APs that should be in range of the RAP backhaul signal to prevent a channel from being overloaded?
I've had an issue with the DFS channel selection flapping. First the mesh will converge using channel 60 then discover radar on that channel. The APs will then move to channel 157 and take about 10 minutes to rebuild the mesh. After using channel 157 for a while the controller decides for some reason to go back to channel 60 only to find radar again and move back to channel 157. Everytime there is a channel move mesh convergence takes roughly 10 minutes. Is this normal DFS behavior?
Has anyone else had problems with this or can shed some light on what's happening?