Cisco Cisco ASA for Nexus 1000V Series Switch Troubleshooting Guide

Page of 4
The MTU of the ASA interfaces must be increased with the mtu command in interface
sub−configuration mode so that the ASA will transmit jumbo frames.
• 
The ASA must be configured to adjust the TCP MSS for TCP connections to a higher value than the
default. If this is not done, ethernet frames containing TCP data will not be larger than 1500 bytes.
The TCP MSS should be adjusted to 120 bytes less than the lowest setting for the interface MTU. If
the interface MTU is 9216, then the MSS should be configured to 9096. This can be done with the
sysopt connection tcpmss command.
• 
What if the ASA is not configured for jumbo frames and it
receives a jumbo frame?
The jumbo frame−reservation command allows not only the transmission of jumbos, but also the reception.
Without jumbo frame support enabled, the ASA will drop packets that are too large. These drops are counted
under the "giant" statistic in the show interface output:
ASA# show interface
Interface GigabitEthernet0/0 "inside", is up, line protocol is up
  Hardware is bcm56801 rev 01, BW 1000 Mbps, DLY 10 usec
        Auto−Duplex(Full−duplex), Auto−Speed(1000 Mbps)
        Input flow control is unsupported, output flow control is on
        MAC address 5475.d029.8916, MTU 1500
        IP address 10.36.29.1, subnet mask 255.255.0.0
        499 packets input, 52146 bytes, 0 no buffer
        Received 63 broadcasts, 0 runts, 5 giants            <−−−− HERE
What if the ASA successfully receives a jumbo frame but
attempts to send it out an interface with a lower MTU?
In order to receive a jumbo frame, the ASA must have the jumbo−frame reservation command, but does not
necessarily need to have the MTU increased (because that only affects the maximum transmission size for the
interface, not the reception).
If the ASA successfully receives a jumbo frame, but that frame is then too large to transmit out the egress
interface, these situations can occur depending on the setting of the Don't Fragment (DF) bit in the IP header
of the packet:
If the DF bit is set in the IP header, the ASA will drop the packet and send a ICMP type 3 code 4
message back to the sender.
• 
If the DF bit is not set, the ASA will fragment the packet and transmit the fragments out the egress
interface.
• 
This is an ASA CLI session that utilizes packet captures to show the ASA receiving a jumbo frame on the
inside interface (with a size of 4014 bytes) that is too large to transmit out egress interface (the outside has a
MTU of 1500). In this case the DF bit is not set in the IP header. The packet is fragmented on egress out
the outside interface:
ASA# show cap in detail
20 packets captured
1: 11:30:30.308913 0017.0f17.af80 5475.d029.8916 0x0800 4014: 10.99.103.6 > 10.23.124.1:
 icmp: echo request (ttl 255, id 48872)
   2: 11:30:30.309920 5475.d029.8916 0017.0f17.af80 0x0800 1514: 10.23.124.1 > 10.99.103.6:
 icmp: echo reply (wrong icmp csum) (frag 48872:1480@0+) (ttl 255) 
   3: 11:30:30.309935 5475.d029.8916 0017.0f17.af80 0x0800 1514: 10.23.124.1 > 10.99.103.6: