>>> On 2/3/2012 at 10:13 AM, Ron Foster at Baldor-IS <RFoster@bald...> wrote:
> 3. You may ask why we used qeth-bus-ccw-0.0.0700 instead of the short
> interface name eth0. The reason why is that in our experience, the short
> interface names
> have this habit of changing. We would come in some Saturday night, take all
> the Linux systems down, apply some z/vm maintenance, and then bring up our
> systems. At that point, some number of our Linux systems no longer
> communicated with the outside world. I finally got tired of that and changed
> all the default
> routes to use the long names.
That sounds very odd. SLES10 uses udev to control that, and the rules are based on virtual device number to make sure the same name gets assigned every time. Was this system upgraded from SLES9 before? That's the only thing I could think of that might have an influence on this since SLES9's use of udev was very much limited and different.
> 4. You might say that that problem is fixed and does not happen anymore. On
> this system I am upgrading the interface we want to use for communication
> to the outside world is eth4. I start out the upgrade, and by the time I
> get to the second part of the upgrade, that is the part where I lose network
> connectivity. The interface name has changed to eth0 and it looks like
> SLES11 SP1 is no longer honoring the long interface name, so this Linux
> cannot communicate with the outside world during the second part of the
> I suppose my next step is to put the SLES10 SP4 system back, and change the
> default route to use eth4 and see if when the upgrade process changes the
> name, it also changes the default route.
I would try just leaving the interface name off entirely, since the kernel should be able to figure out what interface to use based on the IP address of the default gateway.
> Does anyone know where some documentation that documents this behavior?
No, since it's typically not supposed to act this way.