Good day all. We recently upgraded all of our tower routers to 5.1 Our network consists of about 25 towers, And is all OSPF routed internally. And we speak BGP to external peers. All is well with this
We wanted to enable MPLS on our network and did so. We did the following steps.
1. Enabled MPLS package.
2. Enabled MPLS, and filled out LSR ID to be a unique ID for each router (We left everything else default/unchecked on this box)
3. Under the LDP interface tab, We added each Interface that needed to run MPLS. And for each one we filled out the transport address of that local interface. We checked accept Dynamic Neighbors.

We did that on every tower and interface that needed it. And it was working great. (This was last night)
We started getting calls this morning about people not being able to get to some websites (Basically, They could get to google and some other pages, But most didn't work). We also had problems with a few EOIP tunnels that were traversing this network (We would like to upgrade them to VPLS) Bandwidth tests wouldn't run from endpoint to endpoint. And just other strange things. So we disabed MPLS on the network and everything started working again.

What did we miss about MPLS that made this go haywire?

What did we miss about MPLS that made this go haywire?

MTU.

Assuming each packet only has a single MPLS label (4 bytes) then all of your intermediary devices need to be able to support at a minimum 1504 byte packet sizes at layer 2.

On MikroTik you need to adjust the l2mtu parameter, and Cisco its mtu under interface configuration mode. On Cisco you would need to statically set ip mtu to 1500 bytes otherwise IP packets will also start using the set mtu value.

So, If I understand correctly. The L2MTU setting is how you tell mikrotik the mtu of the underlying link?

So if all of our gear was set for 1500 we should set the L2MTU to 1500? And this would tell Mikrotik that the BH is only going to pass things up to 1500?

Let me know if I'm completely missing the point here.

So, If I understand correctly. The L2MTU setting is how you tell mikrotik the mtu of the underlying link?

L2MTU is how you configure the largest layer 2 packet size an interface will accept, minus the Ethernet header. The first picture in that link explains fairly well. You could possibly look at it as a way to signal to MT the MTU of the backhaul, but that really wouldn't be correct. The L2MTU parameter is local to the MikroTik. When you configure it you're telling the MikroTik to accept layer 2 packets up to L2MTU size.

So if all of our gear was set for 1500 we should set the L2MTU to 1500? And this would tell Mikrotik that the BH is only going to pass things up to 1500?

Yes, that's one way of handling it. It definitely makes your config self-documenting and helps others to quickly understand what a link will pass.

I choose to avoid this and keep L2 MTU into in external documentation. MikroTik interfaces 'flap' when changing the L2MTU value. Cisco devices require a reboot. In order to avoid continually disrupting my network when I want to raise the MTU I just set each device to its maximum value. I then document any intermediary devices which use a lower MTU, and take care to not perform any tunneling which will push me over my limit. That process works well enough for me.