We are trying to make use of many 450G’s and external links to create a layer 3 mesh and are having NO success. (The purposes of our meshing are mobility and redundancy.) Please help us understand how to properly configure MME to enable this solution. (Yes, we have carefully studied the 2 wiki topics to no avail.) (In the final solution the wired ethernet will be replaced with external radio links.)
On our test bench we configured three 450G’s to use MME. Each 450G had eth1 configured as an IP in a class C. Eth2, eth3, eth4, and eth5 were each configured with independent, separate, unique class C subnets. Eth2, eth3, eth4, and eth5 were added as MME interfaces. The eth1 Class C was added as a network into MME.
The test bed topology was as follows. All eth2 interfaces were connected to a unique dumb switch. All eth3 interfaces were connected to a different unique dumb switch. All eth4 interfaces were connected to yet another different unique dumb switch. All eth5 interfaces were connected to yet another separate unique dumb switch.
Eth1 on a 450G was connected directly to a laptop with an IP in its respective class C and the laptop was connected to the 450G with Winbox. Eth1 on a 450G was connected directly to another unique laptop with an IP in its respective class C and the laptop was connected to the 450G with Winbox.
The routing tables on both 450G’s (displayed via Winbox) showed that the 450G’s had learned about the class C network on eth1 on the opposite 450G. At this point one laptop was able to ping the other. 
Next we unplugged eth2 from one of the 450G’s and pings stopped. 
For three pings “request timed out” was displayed and then “destination unreachable” was displayed. We were expecting that traffic would be transmitted over either: eth3, eth4 or eth5 which were still directly connected to the dumb switches or that the traffic would be routed via the other 450G which was also connected.
After one minute pings did resume and were traveling direct from the first 450G to the second 450G via eth3. (The MME hold down timer was set to 1 minute).
An additional thing we noticed on the interfaces while MME was running was that there was no dynamic IPIP interface as the wiki explained there should be. This leads us to believe either we are doing something wrong or MME is not functioning correctly. We are using RouterOS 4.7 (the latest stable at this time).
Please help!
