Jumbo frames (w/ df.bit =1) on VSL link?

[Gustavo Novais]

Hello,

We are testing out a VSS solution to a customer who, against our STRONG recommendation will need to plug single-homed some servers into a pair of 6748-GE-TX that exist on VSS both chassis. The distance between the two boxes will be superior to 100m.

We are testing out the scenario where we have jumbo frames crossing the VSL between the two chassis, and we can't get past 1500byte frames.

I do understand and agree that the power of the VSS is on the MEC, but in case some machines are single-homed, and they send out unfragmentable  jumbo frames to the network, they do not seem to get past the VSL link.

The IOS cli does not allow us to set the MTU on the VSL link.

Does anyone have any idea on how to achieve this?

Thanks

Gustavo

[Sergei Vasilenko]

Hello Gustavo,

There is no way to configure the MTU on the VSL (port-channel or physical links).
By default the VSL supports an MTU of 9126 already, so it doesn't require additional configuration.
This is why it cannot be configured from the IOS CLI.

There is one known cosmetic issue with the above that will be fixed in
future IOS versions in the SXI train. When running the command 'sh
interface' the MTU will be showing 1500, however this is incorrect.
This cosmetic issue is reported in the following bug:

CSCsz01779 "MTU on VSL should be 9216":
http://tools.cisco.com/Support/BugToolKit/search/getBugDetails.do?method=fetchBugDetails&bugId=CSCsz01779

So if the frames are actually not passing end2end in your case the reason could be not on VSL link,

but on some other port in transit.

It could on Servers side as well.

Thanks,

Sergey

[Emeric Izsak]

Just a quick note :

Indeed on 12.2(33)SXI4a the "cosmetic" issue has been solved, "sh int Po1" shows correct MTU values but "sh int mtu" still shows 1500.