The MTU of the bodily interface will all the time must be bigger than the MTU of any L3 logical sub-interfaces. That is as a result of when sending an IP packet from a sub-interface the bodily port may also have to append a VLAN tag (4 bytes) to the ethernet header in order that visitors from every sub-interface will be differentiated. Now, as a result of the VLAN-tag is within the ethernet header and never the IP header, it is not counted within the IP MTU of the sub-interface, however it’s counted within the bodily MTU of the mother or father interface.
No less than that is the way it works in Junos OS. It has been a very long time since I’ve wrestled with IOS, however do I recall the way in which MTU is calculated being totally different and never notably intuitive by comparability.
May this trigger points with different routers or implementations? Have
you encountered any implementation the place the sub-interface MTU dimension
adjustments when the bodily interface MTU is modified?
In Junos OS if the sub-interface MTU dimension is not configured, it’s going to change in lockstep with the bodily interface e.g.: in case you improve the bodily port mtu by 10 bytes, the mtu of the logical sub-interface will improve by 10 bytes (however nonetheless all the time be 4 bytes lower than the bodily interface).
If nevertheless, you configure the IP MTU of the sub-interface after which improve the MTU of the interface, it’s going to stay as per your setting, which is a wise default (in any other case you would be overriding configuration which isn’t good).
Should you attempt to decrease the interface MTU or increase the sub-interface MTU to inside 4 bytes of the opposite, Junos OS will throw a commit error and never let you proceed.
The one cause to ever mess with the IP MTU in an OSPF state of affairs is in case you have mis-matched interfaces – ADSL WAN circuits with PPPoEoA at one finish (MTU 1492) speaking to routers with Ethernet (MTU 1500) on the opposite finish is a basic.
On this state of affairs, Cisco has a knob ip ospf mtu-ignore which is able to permit OSPF to type an adjacency whatever the variations. You’ll then want to regulate section sizes of TCP visitors transiting the hyperlink with ip tcp adjust-mss 1452.