12

According to this article two routers will get stuck in the exstart adjacency state when their configured MTU sizes do not match. Does a VLAN tag affect this as well even if the routers involved are unconcerned with which VLAN the packet is coming from or heading to?

I'll try to clarify the question if it is unclear at all.

8

According to this article two routers will get stuck in the exstart adjacency state when their configured MTU sizes do not match. Does a VLAN tag affect this as well even if the routers involved are unconcerned with which VLAN the packet is coming from or heading to?

Most Cisco routers and switches allow 1500 byte IP payloads by default, even when tagged with dot1q. You can verify this with show ip interface SomeIntfName...

Router1#sh ip int vlan105
Vlan105 is up, line protocol is up
  Internet address is 10.15.2.19/30
  Broadcast address is 255.255.255.255
  Address determined by setup command
  MTU is 1500 bytes  <-------------

The reason dot1q on a Cisco interface works without bumping the physical interface MTU is because most Cisco routers support what Cisco calls "baby giants"... a baby giant has an Ethernet MTU above 1518 (which includes the eth header size), but not very much over 1518... usually the default values is 1522 bytes... see this MTU configuration doc for more information. Many Cisco platforms support configurable ethernet baby giant MTUs to 1532 bytes (or even higher), which is also why MPLS tag stacks (2 or 3 tags deep) can work through legacy ethernet links.

Even though Cisco ethernet interfaces support baby giants, they leave the default IP MTU at 1500 bytes. As long as the default IP MTU of 1500 is maintained, you'll have no problems with the default settings of any other router with a 1500-byte IP MTU... quoting RFC 2328 Section 10.8:

10.8.  Sending Database Description Packets

    This section describes how Database Description Packets are sent
    to a neighbor. The Database Description packet's Interface MTU
    field is set to the size of the largest IP datagram that can be
    sent out the sending interface, without fragmentation.
| improve this answer | |
  • What if I am using Cisco router and another router from a different vendor? The Cisco router's interface is configured for a 1500 byte MTU, but it is actually transmitting an IP packet that is 1522 bytes the other router will still accept it as long as the interface is configured with a 1500 byte MTU? That would be because of what @Puglet mentioned below and OSPF assigns the Interface MTU part of the DB Description packet with the MTU size explicitly configured on the interface and not what the actual size of the packet is? – JDGray Jul 31 '13 at 23:18
  • Correct, Cisco IOS uses the IP MTU of the local interface... the IP MTU of your Cisco and the other vendor's box must match. Sadly, I can't speak for the behavior of other vendor equipment. Is there a way you can conduct a ping test (using the DF option) before you deploy? – Mike Pennington Jul 31 '13 at 23:22
  • I'm not actually experiencing this issue. I saw a similar question on link in a discussion about good interview questions. I figured it would be discussion for this Stack Exchange. – JDGray Jul 31 '13 at 23:25
5

So the VLAN tag affects the amount you can transfer through the interface below the Ethernet header, but it doesn't change the actual MTU.

OSPF starts by sending out Hello packets on the broadcast medium (multicast 224.0.0.5) to find neighbours, then performs the DR and BDR election.

After the election has completed, each hosts sends out a DB description packet. Within this packet there is a field [IP -> OSPF -> OSPF DB-Description -> Interface MTU] that OSPF sets to the outgoing interface's MTU.

If peers' MTUs don't match, they won't move past the ExStart phase.

I would take a packet capture (if possible) and drill down into this field to see what's happening, or perform a debug ip ospf adj just to make sure this is the case..

A show interface <int> | MTU will give you the MTU of an interface.

| improve this answer | |
  • So, even if you are adding VLAN tags to a packet it won't affect the size of the OSPF packets negotiating neighbor adjacency because it wouldn't include a VLAN in the first place? – JDGray Jul 31 '13 at 23:21
  • It won't affect the outgoing interface's MTU, which is what OSPF places in its DBD packets. – Puglet Aug 1 '13 at 2:51
2

The IEEE 802.3ac standard increased the maximum Ethernet frame size from 1518 bytes to 1522 bytes to accommodate the four-byte VLAN tag. Some network devices that do not support the larger frame size will process the frame successfully but may report them as a "baby giant" anomalies.

So, if you device's support "IEEE 802.3ac", IP MTU will not change.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.