[F10-nsp] S50s

[David Diaz]

ok there was a known issue with 2.1.5 (old code). On the backside XFPs
(i think it was local to those) I created a LAG.  If packets were sent
through on the front side gig ports at the max size, and u did blan
tagging and sent them over the lag (I think 4bits were added) the
packets would be dropped.  It was a very specific problem and it took
all of about 15secs to adjust the max MTU size allowed on the lag.  I
believe adding a tag put another 4bits on and that pushed it over the
default limit in some cases.

2.1.6 corrected the limit be default. That is the single only problem
I have run into and that was because I was an early adopter.  The
boxes have been hammered on and have stood up just fine.  I am soon to
try 2.2.1. I know the CLI syntax has changed a lot.

As for crashes, reboots, phantom MAC addressing etc showing up, zip,
and this has been an issue with other vendors. I would compare force10
to juniper but just in the ethernet space.

It would be interesting to see if anyone has stacked 8 units and
banged on all 384 ports to see what kinda stats they produced.

Oh the important issue with the bug, TAC verified and diagnosed it in
about 10 seconds.  We came up with a solution immediately that would
require no interruption of services... non reboot.

dave



On 2/16/06, Andy Myers <acm at dullroar.org> wrote:
> On Wed, 2006-02-15 at 22:04 -0500, David Diaz wrote:
> > I saw a minor MTU problem on the LAGs but besides that it has be rock
> > solid and I also found the support very good.
>
> Hi, I think I was the one who was originally asking about S50
> experience...  We're planning to use LAGs too, so could you elaborate
> about what went wrong and how you fixed it?
>
> Thanks,
> Andy
>
>
>