[c-nsp] Cisco 1841 Router CRC errors

[Higham, Josh]

> [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of Tom Storey
> 
> On 18/12/2007, at 7:42 AM, Paul - Talk Talk wrote:
> 
> > We have upgrade our lease line from 4mb copper to 10mb fibre which  
> > meant we
> > had to change our router.
> >
> > The old router was a Cisco 2620 and the new Router is a Cisco 1841.
> >
> > Between the Cisco Router and our Cisco 3550 layer three switch is a
> > bandwidth management box, which is an Allot NetEnforcer AC-402.
> >
> > On the old the old set-up using the Cisco 2620 everything 
> worked fine.
> >
> > On the new set-up with the Cisco 1841 we are seeing a lot of CRC  
> > errors
> > until the whole thing falls over.
> >
> >> From our side nothing has changed except the router which was  
> >> supplied by
> > the ISP.
> >
> > We did find out that the FE0/0 interface on the Cisco 1841 was set  
> > to auto,
> > but this has now been set to Full Duplex 100mb which 
> matches the Allot
> > NetEnforcer.
> >
> > Taking the NetEnforcer out and connecting to the Cisco 1841 
> directly  
> > to the
> > Cisco 3550 stops the errors, but as said before switching back to  
> > the Cisco
> > 2620 also stops the errors.
> >
> 
> On one particular PC I have cannot auto negotiate and cannot 
> be set to  
> 100/full without producing errors. I have to run it in 100 half. At  
> least this one is only a PC. Its got a Broadcom NIC.
> 
> My suggestion would be, if you can spare two ports on the switch, to  
> create a VLAN and stick the two interfaces that would 
> normally connect  
> directly to each other into that VLAN. Otherwise, try 
> different cables  
> between the two, set the port to 100/half, and try different  
> interfaces if you can.
> 
> I may have other issues with the PC above, such as a dodgy cable. I  
> havnt gone so far to test that out yet as its bandwidth requirements  
> are low, 100/half does just fine.

I can't imagine that running half-duplex is a good idea.  I have had
interfaces on the 28xx series (in particular with the HWIC ethernet)
where they have had to be set to auto/auto on both sides to not produce
errors.  This includes connecting to a Cisco switch on the other end.

Since your upstream is 10Mb you can probably also get away with running
10/Full, although it makes increasing bandwidth in the future a bit more
cumbersome.

Thanks,
Josh