This is probably something we should really be raising through TAC but I thought it might be worth seeing if others are running into the problems first. We're running a number of Meru MC4200 controllers with version 8.1-3-2 of the firmware.
The problem in question is actually something we're seeing in the NMS we're working with. Monitoring the bond0 interface, we're seeing some odd behaviour during "high" traffic periods.
We're seeing the interface manage to push 300-400Mbps through on 1Gbps copper. The weird bit is the graph levels out, as if you've hit a bandwidth limit somewhere. It still varies a bit but remains relatively level. As traffic flow tries to increase throughout the peak, we see a peak in RX discards on that interface increase to match the peak. In some cases, we're seeing as much as 500k discarded packets in an hour. That's discards, not failing on CRC error or similar as reported by the controller via. SNMP.
That doesn't sound normal to us. There seems to be no issue in terms of CPU or memory according to the controller. Is this just a performance bottleneck we need to work with in these controllers?
