Is there a bug in MPLS and older hardware?

So Ive been pulling my hair out, trying to fix a network that was always working fine on 5.26.
I have older hardware with RB750s and RB433s, RB493s that only support 1522 MTU on some interfaces.
So my MPLS MTU is set to 1522 across the board network wide, but I moved the RB750s to 6.13 or 6.17 and all of the sudden MPLS is broken, I can’t get 1522 bytes through anymore!
But the RB493 with a different switch chip, and the same 1522 MTU on all its interfaces, is unaffected. Newer router boards are also unaffected.
Outside of MPLS, I can set an IP MTU of 1522 between two adjacent routers and get the packet through, so it seems its something in layer 2.5 MPLS only.

EDITed to add, this doesn’t affect RB750s at the end of the MPLS line (VPLS tunnels not passing through the router) If the tunnel is terminated on the 750, its OK, but if the 750 is in the middle with VPLS tunnels passing THROUGH it, the next router in the line is affected.


My MPLS MTU is 1522.. no VLANs, so my largest VPLS unfragmented packet is 1496, I know, and with my PPPoE overhead is 1488. And it works fine on 5.26 on the older hardware.
Am I losing my mind??

I hit something like this using RB2011 and RB750G as a tunnel terminators (2 wireless links between them).
However, i didn’t spend too much time over it because i plan on swapping the RB750G with a 2011.

Did you find something else?

What frame size CAN you get through?

It would be interesting to do a packet capture and look at the frame to see if something is eating up extra header space.

just wanted to say that i replaced the 750g with 750gl and it magically works.
750G and possibly older units have a lower MTU limit than the gl variant.