Problem with Bridge or VPLS?

Ok, so I set up my new network… looks like this:
network extension.png
R1/2/3/4 = RB1200 w/ROS 5.12
AP1/2/3 = RB435 w/ROS 5.12

R1 config:

/interface bridge
  add name=vlan_bridge protocol=rstp

/interface bridge port
  add bridge=vlan_bridge interface=ether6 horizon=1
  add bridge=vlan_bridge interface=vpls-r2 horizon=2
  add bridge=vlan_bridge interface=vpls-r3 horizon=2
  add bridge=vlan_bridge interface=vpls-r4 horizon=2

R2/3/4 config:

/interface bridge
  add name=vlan_bridge protocol=rstp

/interface bridge port
  add bridge=vlan_bridge interface=ether6 horizon=1 
  add bridge=vlan_bridge interface=vpls-r1 horizon=2

AP config:

/interface bridge
  add name=vlan_bridge protocol=rstp

/interface bridge port
  add bridge=vlan_bridge interface=ether1

/interface wireless
  set 0 wds-mode=dynamic wds-default-bridge=vlan_bridge

At this point, everything is up and running. I can attach VLAN interfaces to the bridge on each router and/or AP, assign IP addresses, and I can pass traffic between R1 and the other routers and APs. I can even see VLAN traffic hitting the switch behind R1. It works beautifully.

When we added our first customer, VLAN 1801, the turnup was flawless. Same thing with VLAN 1901 & 1902 off the second AP. Everything seemed to be working without issue.

Then, we got a call from 1801 saying his internet is not working. So, I log into his CPE, stick a DHCP client on the bridge and get a binding (dhcp relay on the switch hanging off R1). I start doing some pings and traceroutes, but I’m getting 50-70% packet loss when it works at all.

I hop over to the 1901 CPE, and it’s working great. Same thing with 1902, working flawlessly.

Back to R1, I disabled the VPLS-R2/R3 bridge ports., and 1801 started working perfectly. Re-enable VPLS-R2, and 1801 started acting up again.

Finally, I removed all VPLS tunnels, all MPLS configurations, and put the VLANs directly on the physical layer. Everything is working again, but this is not sustainable, as I do not want a bridged/switched network (there’s unrelated issue to raw throughput when routing through the RB1200).

So, is this a problem with the bridge or with the VPLS tunnels? I’m thinking bridge, but for the life of me, I don’t know what it would be.

Thanks for any input anyone here can provide.

-Troy

how is your MTU trough your lisenced radioes? I have to manually specify bigger MTU than 1500. and how about AP/client MTU (Might not be the problem, but worth a check) I also dont like RSTP in setups like this. It can make more problem than good. (try connect a bad RSTP configurated switch at your customer site)i cannot see a configuration here. A default RSTP setup, without a specified root bridge, can make problems.