I have a hotspot running on a bridge. There are several sub nets on the bridge. All but one are private ip ranges. Downstream there are several Ubiquity APs. Customer devices are Ubiquity as well and each has a static private ip and the vast majority are setup in router mode. A few have a bridge setup.
The hotspot’s ip is on that same private ip network and it is configured to not do nating (no universal client). Not every subnet on the bridge is subject to the hotspot so all but the one are excluded via the hotspot’s ip binding setup.
The hotspot was configured with radius but at the moment radius is set to accept all requests. We do mac authentication. The radius server was populated with macs and returned the Mikrotik-Rate-Limit attribute so simple dynamic queues are setup.
When the hotspot was activated the vast majority of clients seemed to be ok. Radius authenticated. We could see the hosts in the Active tabs, bandwidth usage, etc. However several clients had very spotty bandwidth. There would be times when they got little or no bandwidth, then other times the client reports all was well (this pattern continued). A few client had no access (these may have reached the shared host limit).
Anyway we also saw lots of public ips listed in the Hosts tab and well as private ips that aren’t even on a bridged subnet. I don’t know if this in itself would cause a problem but reporting it in case someone knows.
We also saw a few mac/ip combos denied by the hotspot when the max (per mac) limit we set (at 4) was reached.
The one last piece of information: prior to configuring the hotspot, we had manually entered queues. Those were not removed (yet) so the radius setup ended up duplicating queues for each connection. I’ve read that long queue lists can slow things down but will this setup cause the performance degradation we saw?
Sorry for the long winded explanation. I can provide more details if needed.
So is there anyone out there that explain what is wrong or suggestions for troubling shooting this?