cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
2749
Views
3
Helpful
2
Replies

Cisco 2960X MTU issue

joel.cason
Level 1
Level 1

I went through an issue earlier this year on a 3850 stack where we were trying to enable jumbo frames on a single L2 VLAN for a UCS environment connected via port channel to upstream switches. If the interfaces are both on the same UCS FI, jumbo works just fine.  But we have issues when trying to cross the upstream switches.  The jumbo enablement seems good (I can send >1500 byte ping), but after 1562 bytes things just die whether or not DF bit is set.

Another person fixed the issue on the 3850 stack which they believe was related to auto QoS.

We are now in a different deployment with a 2960X stack, same hardware configuration for the UCS stuff, and the same exact issue.  I can send a 1562 byte packet with or without DF bit set, but as soon as I go 1563 or larger they drop into nothingness.  I don't get a "fragmentation needed" message on the sender side.

I ran tcpdump between the two hosts and I see the 1562 send and respond on either side.  For the 1563, it sends but no response received anywhere.

Auto QoS I'm told isn't being used here so it would have to be something else. 

Kind of desperate for ideas at this point as to what would be causing the same failure at the exact same 1562/1563 byte breakpoint.

Code on the switches has been updated to the latest and greatest and didn't help this issue.

2 Replies 2

joel.cason
Level 1
Level 1

Can't figure out how to edit the original question.  Meant to add, the person who fixed the 3850 believes it was related to auto qos but can't be sure and of course we have no notes or documentation on it.

egarcial
Cisco Employee
Cisco Employee

Hi Joel,

I have read you current issue and I really dont think that the 3850 issue was fixed  with auto QoS, since this will not modify the maximum packet size that the interface or switch can handle. On the 3850 all the interfaces are QoS trust, this means that the QoS marking on all the packets is allow and did not change when you applied the auto QoS.

At this moment I really think that you problem will be solved if you change the MTU size to 1600 or up to 9000. Basically we need to use the same MTU size on both sides, on the UCS and on the switch.


You can  follow this document and take in consideration the restrictions.

https://www.cisco.com/c/en/us/td/docs/switches/lan/catalyst2960xr/software/15-0_2_EX1/int_hw_components/configuration_guide/b_int_152ex1_2960-xr_cg/b_int_152ex1_2960-xr_cg_chapter_0110.pdf

Restrictions for System MTU When configuring the system MTU values, follow these guidelines:

• The switch does not support the MTU on a per-interface basis

• If you enter the system mtu bytes global configuration command, the command does not take effect on the switch. This command only affects the system MTU size on Fast Ethernet switch ports.

What it takes my attention is that it works with the MTU size of 1562. This since the maximun MTU size that is not consider jumbo frame is 1500 + 26 of the payload. So not sure where the extra bytes are been allowed. A "show interface", "show run interface x/x/x" and also "sh run | in mtu" should give us more information about why the packet is been actually supported.

Also a capture will help to identify the actual size of this frame. But the problem should be fix after change the desire MTU size on both sides.

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: