Symptom/Cause | Solution |
|---|---|
Within MPLS network if the P routers are of other vendors, due to incorrect MTU settings BGP session can flap if services (L2VPN/L3VPN) exchanges exceed the MTU size. | 1. Note that on OcNOS device when MTU is configured on L3 interface, it implies only the payload (till L3 headers). • Additional MTU size is increased in hardware as follows: L2 header : 14/18 bytes MPLS header : Up to 20 bytes if label-switching is enabled on interface 2. So if MTU set is 1500, hardware sets MTU as following: • MTU size in kernel : 1500 (data is fragmented by kernel beyond 1500 bytes) • MTU size in hardware (label-switching not enabled on interface) : 1500 + 14/18(L2 header) • MTU size in hardware (label-switching enabled on interface) : 1500 + 14/18(L2 header) + 20 (Up to 5 labels) 3. So corresponding interop device has to configure MTU considering this increase. 4. Also, it is always recommended to configure MTU higher on the network side interface than the access interface, due to following: • For L2VPN/L3VPN services additional up to 2 - 5 labels can be inserted on network interface. (1 tunnel label, 1 service label, 1 LU label, 2 entropy label) • For L2VPN additional L2 header is inserted, so that has to be also accommodated. |