Strange things on GBit / 1000->100 / net.inet.tcp.inflight.*

[Andre Oppermann]

"Raphael H. Becker" wrote:
> 
> On Sat, Sep 18, 2004 at 01:11:13AM +0200, Andre Oppermann wrote:
> > I'm unable to reproduce your problem.
> >
> > Usually in these cases there is an ethernet duplex mismatch somewhere.
> > Can set the network cards and the switches to fixed 1000 or 100 full-duplex?
> 
> Yes, one of my first tests. Both machines fixed to their nominal
> transfer rate. Tested different ports, different irons (source and
> target), different cables. Always same result.
> 
> All channels are able to do their maximum speed.

Ok.

> Changing target and source (100 -> 1000) resulted in >10MBytes/sec, same
> machines, same cables, same everything 1000 -> 100 resulted in what we
> discuss here (300kBytes).
> 
> 100->100 (between the two PE350) is working fine, too.
> 
> The problematic case (1000 -> 100) was tested with 3 machines 5.x and 2
> machines 4.x, all cases the same.

Hmmm...

> I'm not sure this is a "physical" problem.

The high number of out-of-order packets on your 4.10 machine is very
suspicious.

> I will try something else next days:
> * New cables
> * FreesBIE with 4.10 on the 5.3-Boxes, ftp'in /dev/zero directly to /dev/null ...
> * throtteling 1000MBit ... maybe
> 
> Maybe:
> * another GBit-Switch (if available)

That would be nice.  Try to get some other brand or model than the one
you have now.

> Workaround:
> * Buy 1000MBit-Cards for the both "Router"-boxes

Why that?  Do you have to do TCP transfer to/from these machines?

Yet another test you could try:

 5.3 GE ---> FE 4.10 "Router" FE ---> 4.10 FE

-- 
Andre