Gentoo Box Drops Off Network (Long)

[Gerald Combs]

On Thu, 1 May 2003, Steven Elling wrote:

> Every time the box drops off the network I run 'netstat -i' and every time 
> it shows RX-ERR and RX-OVR in the hundreds.  This in itself makes me think 
> there is a problem somewhere in the network subsystem, but where?  Are 
> there any kernel / module parameters I can use to further trouble-shoot?  I 
> can't find any.  I've done everything but try a different brand of network 
> card.

Overruns indicate that data is arriving from the network before it can be
moved into your computer's memory.  This can be caused by a bad driver, a
crappy chipset (e.g. the RealTek 8139), or both.  (It can also be caused
by your card being "faster" than the computer, but I'm assuming you're not
running GigE or 100 Mbps Ethernet on a 486.)  If too many errors or
overruns accumulate over a given period of time, the card or driver may
drop off the network temporarily.

What chipset and driver are you running?  You indicate that it ran fine
under Debian; you might compare that driver and version to the one you're
using now.  You could also emerge the "mii-diag" package; it can give you
more information than you ever wanted about your Ethernet card.

> I wouldn't think a memory problem would cause it to drop off the network.  
> Plus, I don't believe I have a memory problem because I would see other 
> problems manifest themselves.  For one, I compile with '-O3' and if there 
> were memory problems the compiles would bomb out with 'Seg 11' at least 
> once.  Regardless, I think I am going to run memtest overnight to see what 
> it comes up with.
> 
> What is interesting is the network problem started imediately after I 
> replaced the OS on the machine from Debian to Gentoo.  The way I did this 
> was I installed a spare drive in my workstation; which uses the same 
> processor and motherboard; mounted the drive, unpacked stage1 onto the 
> drive, chrooted into the root of the drive, then proceeded to build Gentoo.  
> Once the build was complete and I had all the necessary software installed, 
> I swapped out the drive in the server that had Debian on it with the Gentoo 
> drive and proceeded to complete the install.  I did a readonly badblocks 
> scan on the Gentoo drive and it didn't turn up anything.  The server never 
> once dropped off the network in the 2+ years that Debian was installed on 
> it.
> 
> So after reading everything above, can anyone make any suggestions as to 
> were I shoud start next on tracking down the server's network problem?
> 
> Thanks.
> 
> 
>