[wellylug] eth troubles

[Abhinav Keswani]

> > check duplex
> 
> Should it be on or off ?
> 
> > and speed settings on the eth interface - is it
> > autonegotiating ok with the switch?

Right.  You should be asking whether its full or half.  :)  Here's a
good place to learn about this - first thing I saw in google:

http://www.tcpipguide.com/free/t_SimplexFullDuplexandHalfDuplexOperation.htm

But to get you under way - you want your interfaces to be in full-duplex mode:

"In full-duplex operation, a connection between two devices is capable
of sending data in both directions simultaneously."

Usually problems exist between switches/routers and network interfaces
in terms of negotiating the duplex type for a connection.  Cisco
routers have notoriously bad history with autonegotiating this.  So
what most people do is to hard set this on both ends of the connection
- to full duplex.  Problems arise when one side of the connection is
full and the other is half duplex - one gets packet collision and
connections get reset and everyone is generally unhappy.

So - check this out by doing:

[root at wasabi root]# ifconfig -a
eth0      Link encap:Ethernet  HWaddr 00:0B:CD:48:52:2E
          inet addr:192.168.16.150  Bcast:192.168.23.255  Mask:255.255.248.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:77104619 errors:0 dropped:0 overruns:0 frame:0
          TX packets:51231050 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:100
          RX bytes:1616724456 (1541.8 Mb)  TX bytes:53032679 (50.5 Mb)
          Interrupt:5 Base address:0x1000 Memory:fc500000-fc500038

Here you can see I have no collisions reported, no errors, no packets
dropped...all good.  Do you have the same?

Then run something like this:

[root at wasabi root]# mii-tool
eth0: negotiated 100baseTx-FD, link ok

Which tells me that my luverly machine has a NIC up at 100Mbps Full-Duplex.

mii-tool comes with most distros - you'll need to install it if you
don't have it.

Ideally your output for this should match the setting on the switch. 
Can you do any diagnostics on the switch to see how the ports are set
up?

> > can you check for errors on the switch port that this host is connected to?
> 
> Not sure I have been given it and I do not have a manual
> 
> > try clearing the arp cache on the switch and the host
> 
> Where is it on the host ? never done that before...

Do this to clear your ARP cache:

First check to see if anything is in the arp table:

arp -a

Then selectively remove an arp table entry that corresponds to your switch by:

arp -d <hostname>

Ideally you would be able to do this on the switch as well.  I don't
know the switch you have and if you don't have the manual I can only
recommend you look up an online manual or google it.  Power cycling it
should drop the arp cache ...  perhaps why it worked the first time? 
Dont know why it failed the second time.

By doing the above two things you're:

1. seeing if there are errors in the connection.
2. establishing a clean slate for the switch to reinitialise itself.

-Abhinav