If you are running Layer 2 code on your Brocade FastIron then all you need to do is enable jumbo by;
Once the switch reboots the switch will be able to pass jumbo frames. You will then need to enable jumbo frames on the NIC’s of the clients that require jumbo frame support.
What happens when a client sends jumbo frames to a client that is not set for jumbo frames at Layer 2?
This depends on how the client is sending the datagram;
TCP/P/Ethernet : There is a setup process for each session between sender and receiver where devices exchange their maximum MSS (Maximum Segment Size) value. The lowest value MSS determines the MSS used for that session, and the subsequent MTU and frame size.
UDP/IP/Ethernet : The packet is dropped by the receiving client.
Setting MTU values for Layer 3
First you must enable layer 2 jumbo frame support or you will receive an error setting the MTU. If you see the below result then you have not enabled Layer 2 jumbo support.
“Error - Invalid input 9000. Valid range is between 576 and 1500”
To set the MTU for a Layer 3 VLAN
The VE (Virtual Ethernet) must be already defined on the VLAN.
FastIron(config)#interface VE 14
FastIron(config-vif-14)#ip mtu 9000
To set the MTU in Layer 3 for a port
The port must be a layer 3 port and not a switch (layer 2) port, this means the port should have an IP address assigned to it. Choose a port that is a member of the default VLAN (default VLAN is 1)
Use ‘Show VLAN 1’ to find a unused port.
FastIron(config)#show vlan 1
Total PORT-VLAN entries: 6
Maximum PORT-VLAN entries: 64
PORT-VLAN 1, Name DEFAULT-VLAN, Priority level0, Spanning tree On
Untagged Ports: (S12) 16 17 18 19 21 22 23 24
Tagged Ports: None
Uplink Ports: None
DualMode Ports: None
Let’s make port 12/24 a layer 3 port and assign the MTU value of 9000.
What happens when a client sends jumbo frames to a client that is not set for jumbo frames at Layer 3?
Now when a jumbo packet is to be routed, it is each routers job to deliver the packet to the next hop. So if a jumbo packet is received on a routers ingress port that is jumbo enbaled and must egressed on a port that is not jumbo enabled, then the router will fragment the packets into chunks that the egress port can handle. Note when the router must fragment packet this will be done on the CPU of the router.
However if the packet has the DF (Do not Fragment) bit set, then the router will send a ICMP message of “fragmentation needed and DF set”. This will be routed back to the sender so it can lower its MSS. Note that I lot of firewalls will block ICMP which will lead to black holing.