I have seen information on this problem all over the Forum. However I have not experienced it until recently.
I have multiple AP - Blend of RB912U N and Net Metal RB921 AC
We are in the process of moving to all Net-metal RB921 AC.
On all systems we are seeing a dramatic decrease in TCP traffic. Testing VIA IPERF see below
LINUX/IPERF---->RB911CPE------->RB912 AP---->RP1100—>Linux/IPERF
We are seeing stable DL from AP to CPE however we are seeing almost nothing from CPE to AP. Moreover the AP disconnects while running the upload test.
At first i thought it might be a weak power supply but all Power Supplies we have are 24v 1 amp.
I have no idea what is causing the CPE disconnect during the upload tests. Now this is not an isolated incident. It is happening across everything. We are using ROS 6.18 and 6.27 Mind you the APwith 6.18 has all its CPE’s on 6.18 and a AP with 6.27 has all its CPE’s on 6.27.
I then thought buffer bloat but i can’t find anything on it.
Does anyone know of any issues with CPE to AP upload problems with 6.18 or higher.
Additionally we are using NV2 we see no difference if moving from NV2 to nstream or 802.11
Final Thought Could it be a queue problem. We are using wireless-default for Wireless and only-hardware-queue for Ethernet.