I'm testing next scheme for over month:
Problem is thet this setup absolutely unstable.
I'm trying to transport user VLANs over mesh network: from client to gateway.
Trying to test mesh self recovery occurred next things :
Try disable one of the mesh nodes:
1. disable enable mesh portal - like electricity outage. mesh recovers, can ping each mesh node, but users can't ping each other/gateway anymore. Never recovered until restart whole scheme one by one.
2. Disable one of the client access mesh nodes B or C. Enable; mesh recovers, can ping each mesh node. Users stop pinging each other or gateway. Same, never recovers until restart.
Sometimes this occurs while disabling only one transport link on one of the mesh nodes. Transport link which transports data in moment of disable.
Very rarely setup recovers without problem. Never recovered after disabling whole mesh node (electricity outage).
Tested on ros 5rc10, 5rc11, 5, 5.1
After final ros 5 can't setup admin mac on mesh interface/bridge. I could do it in ros RC's but system stopped working at all after enabling admin mac. Now winbox shows me error messages like "timeout cant change interface" or "unable to change interface" (sry don't remember error messages exactly)
Also tried putting mesh ethernets in separate bridges and after all putting them into mesh bridge (as suggested in wiki "for stability")
In RoS 5 mikrotik has fixed error regarding problem of creation VLANS on mesh/bridge/bonding interfaces http://forum.mikrotik.com/viewtopic.php?f=1&t=49993, but looks like they just have fixed letter I (invalid) in this error and not actual problem of disabled VLAN interfaces .
In my opinion while mesh recovery some loop occurs in Vlans ....whatever don't know what could be problem else.
Tested for one month everything that could be configured and implemented in HWMP+ mesh network...