Loosing Connection To subnet

[Tom Mitchell]

On Sun, 14 Dec 2003, Patrick Nelson wrote:
> On Fri, 2003-12-12 at 15:01, McKeever Chris wrote:
> > have this strange intermittent issue, on two separate
> > occasions, on the same box spanning 2 entire scratch rebuilds
> > of this machine, I can not ping a certain subnet, other than
> > the router.
...
> > if I restart the network service on the computer having this
> > problem, it all comes back online.  I can however during the
> > whole time reach the subnet and computers on it without a
> > problem from any other machine.
...
> what nic and driver is in use?  I have a very similar intermittent
> problems.  The nic looks like it's up and running, but it isn't sending
> anything.
....
> time).  It drives me crazy!

Is there one or more  routing daemons active (can only be one)?
  (zebra, routed, ripd, ospfd, ipx_route radvd, rdisc)
  There was a recent update to zebra... hmmmm.
If one daemon, how is it configured?
Is there a default router that you are using, what is it and how
is it configured?   A good default route and you do not want
or need a routing daemon.  Host routes?

Double check broadcast addresses!

It is possible for initial gateway and routing info to be steped
on by a routing daemon that thinks it knows what is going on.  
With bad routing protocol interaction, things work, get quiet and
stop working.

Default host names and addresses can be "corrected" some time
after you boot.  If DNS/NIS data is wrong (mis matched) things
break after the service is established and cached values expire.

Check for unexpected broken hard/soft links on host files and
other network config files (hosts, ifcfg-eth0,resolv.conf, ...)
$ ls -li /etc/hosts /etc/sysconfig/networking/profiles/default/hosts
6734009 -rw-r--r--    2 root     root         2062 Dec 12 12:51 /etc/hosts
6734009 -rw-r--r--    2 root     root         2062 Dec 12 12:51 /etc/sysconfig/networking/profiles/default/hosts
# ls -li /etc/sysconfig/networking/devices/ifcfg-eth0 /etc/sysconfig/networking/profiles/default/ifcfg-eth0
3604564 -rwxr-xr-x    2 root     root          173 Nov 22 18:40 /etc/sysconfig/networking/devices/ifcfg-eth0
3604564 -rwxr-xr-x    2 root     root          173 Nov 22 18:40 /etc/sysconfig/networking/profiles/default/ifcfg-eth0

Run arpwatch on a couple machines.
   (some NICs work better/ differently in promiscuous mode ;-)
   (arpwatch may tell you stuff of interest).

Snoop for routing packets with tethereal/ethereal or tcpdump.
Also, inspect packets that do not belong on the net.

Use traceroute/xtraceroute when things work and make sure that
they work as expected.

Some NIC setups can set MAC addresses that conflict if you copy
files about or swap hardware from box to box. In general one
should leave MAC addresses to factory set values.  Ether switches
will hurl on MAC addr conflicts.

-- 
	T o m  M i t c h e l l
	mitch48 -a*t- yahoo-dot-com