[c-nsp] Jumbo frames on certain VLANs with UCS fabric?

Manu Chao linux.yahoo at gmail.com
Tue Nov 16 12:53:03 EST 2010


Hi David,

(L2) MTU is not per VLAN but per physical interface.

Make sure all interfaces in a VLAN are configured for jumbo frames before
configuring jumbo frame support on an SVI.

You have to enable jumbo on ALL your 4900 switches interfaces.

You have to enable jumbo on your UCS (The MTU is set on a per CoS basis in
UCS).

R/
Manu

On Tue, Nov 16, 2010 at 5:19 PM, David Hubbard <
dhubbard at dino.hostasaurus.com> wrote:

> Hi all, I'm working on deploying a UCS system using
> iscsi to an EMC with only a pair of 4900M's in
> between.  I'm having a bit of trouble wrapping my
> head around what I need to do to enable jumbo frames
> to make it from end to end on the storage vlan.
>
> What I've got so far:
>
> 1) Two redundant UCS fabric interconnects are each
> in end host mode and each dual-home to the two 4900M's
> at ten gig.
>
> 2) Within the UCS configuration, there is a virtual
> nic defined as mtu 9000, the rest of the virtual nics
> are defined as mtu 1500.
>
> 3) The 4900M's have a storage vlan configured with an
> mtu of 9000.  All other vlan's are default 1500.
>
> 4) The ten gig interfaces to the UCS are set as mode
> trunk but not mtu 9000 (I think this is an issue?).
>
> 5) The ten gig interfaces to the EMC's two service
> processors on each 4900M are configured as
> 'switchport access vlan...' and 'mtu 9000'.
>
> 6) The two 4900M's have a trunk between them carrying
> all vlan's.  (This may be an issue too?)
>
>
> show int on the interfaces to the UCS fabrics and the
> switch to switch trunk are both showing mtu 1500.  Do
> I need to set all of those interfaces to 9000?  I tried
> that, but did not test the EMC connectivity in between
> because I noticed as soon as I set that I got a bunch
> of MTU_Mismatch showing in "show vlan mtu" and I was
> worried I was about to create a big problem.  Or, is
> it ok to have the mismatch there because any device
> talking on the 1500 byte vlans would not generate a
> larger packet anyway but devices on the 9000 byte vlan
> are free to talk jumbo and the interfaces will transport
> it because it's mtu 9000 all the way through, including
> the switch to switch trunk?
>
> Thanks,
>
> David
>
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>


More information about the cisco-nsp mailing list