Hi
I am having a problem which is affecting some of our sites where we use Mikrotiks.
This is the setup:
Routerboard RB1100AHX2 with multiple pppoe uplinks (ADLS, through modems in bridge mode).
Hotspot configured on the interface facing the customers.
In order to load-balance between all pppoe interfaces we have created a 0.0.0.0/0 route with all pppoe interfaces (ecmp), and masquerade the customers.
All pppoe interfaces have the same gateway IP address, as they are form the same provider.
Therefore only one /32 route is created for the gateway’s IP address, and one pref-source selected for that route.
The problem is that the gateway drops packets with source IP different than the one it assigned to that ppp tunnel (understandable). For example pppoe1 should have a source IP
of 1.1.1.1, but the Mikrotik sends what is sees in the pref-source of the routing table for that route (1.1.1.2).
In my opinion the Mikrotik should change the source IP address of the packets before sending them out of the interface, and not using the same source IP across all ppp connections.
To resolve this I have applied the following rules:
add action=src-nat chain=srcnat disabled=no out-interface=pppoe-port1 to-addresses=1.1.1.1
add action=src-nat chain=srcnat disabled=no out-interface=pppoe-port2 to-addresses=1.1.1.2
add action=src-nat chain=srcnat disabled=no out-interface=pppoe-port3 to-addresses=1.1.1.3
one for each ppp connection.
After that was applied DNS started to be resolved again and Radius auth stopped being dropped too.
The problem is that we can only access the Mikrotik from remote using the IP of the pppe connection which has the IP in the pref-source for the gateway (which changes when the ppp connections are restarted), and our monitoring reports the IPs are unreachable (ping is not coming back).
The behavior is the same with firmware 5.2x and 6.0
How can I make sure pakets leaving the pppoe tunnels have the right IP address for that interface?
Am I missing something/doing something wrong here?
Thanks