I have encountered an unexpected problem in a switched network, where it seems to be possible with only two switches between a client and a ROS Hotspot.
This works fine:
client–switchA–switchB–RB_hotspot
This does not work:
client–switchA–switchB–switchC–RB_hotspot
The symptom is that with three switches we can see the URL the client is requesting coming in to the RB hotspot (web-proxy log), but the logon page is never received by the client. The browser just keeps spinning waiting for a response.
We have tried swapping switchB and switchC, and used different switch brands, but always get the same result..
Before tearing the complete architecture apart, is this a real limitation and is there a logical explanation?
There is no explanation for this behavior. Please post a more detailed description of your setup (switches, version of ROS, is it x86 platform or RB).
Also try a bandwidth udp test to see if there is a bottleneck between the client and ROS Hotspot.
Yes, all switches are manageable. I’m not sure what you mean with a flat mode, but we use many different vlans.
In the router the hotspot is configured on a vlan interface, and this vlan is then tagged over all switches. The switch ports on the final switch where cpe’s are connected, are configured as default for this vlan and untag on egress.
Going to have to tell you to go back to the switches and poor over their configuration and make sure everything matches. This is especially true when you are dealing with VLANs, if the ports aren’t tagged properly between the switches, VLANs won’t go over the switches, or if the port to the MT isn’t tagged properly it can’t reach it. Since when you add in “switch-c” and things break, I’m assuming it’s the configuration on that switch that is the problem.
As was suggested, it would be useful for you to supply a network diagram as well as an /export of your current configuration with the relevant parts of it.
/interface export
/ip hotspot export
/ip route print detail
/ip address print detail
All, thanks for your response!
My initial question was just if this was a known problem. Since it’s apparently not, I agree with you that the issue should be in the switches downstream somewhere. ROS and hotspot configuration is unlikely to be an issue since hotspot works just fine with one and two switches between the RB and the CPE. On the other hand everything works fine also on the third switch if hotspot is disabled in ROS, which makes it a bit mysterious.
We will start by taking this back to our lab and replace/reconfigure the switches involved.