I am slowly coming to the realization …

that basic systems management and configuration are neither well known nor well understood, by the vast majority of people. Even the ones with certifications who should know this stuff.
This is leading me to rethink some of the basic elements of the out-of-box experience. Part of this is driven by the tendency of certain distributions to remap, effectively randomly, their network port assignments. This is never, under any circumstances, a good thing.
Unfortunately, this happens. All you need is a quick yum, and whammo … eth0 is now eth3 and vice versa. Let me re-iterate. There is no conceivable universe where this is a good thing. It stresses people unboxing units and updating software far beyond their limits, and wastes time for installations.
Its time to really step up with our config tools. SMASH is coming along nicely, so we will probably start shipping it soon with the systems. This should alleviate … oh … 99% of problems with configuration.

6 thoughts on “I am slowly coming to the realization …”

  1. I have repeatedly yelled myself hoarse at all those involved (OS and hardware) about this. It’s always the other guy’s fault. And none of their solutions scale like “the first port is ALWAYS eth0…”

  2. I should note that I previously worked as the Global Linux Engineering lead for a company with over 15,000 Linux boxes. So I _thought_ I had some clout.. silly me.

  3. This one drives me nuts too . . . and I work where that sausage is made. What really amazes me is that if you look in the per-interface config files they even have MAC addresses associated with interface names – so some script has to be deliberately editing those files to mess things up. Insanity.

  4. Once upon a time, device enumeration order was not deterministic (over a flat bus). OSes left renaming / renumbering to user-space programs that compare MAC addresses (for network cards) to a listing. In that context, the solution makes sense.
    That legacy has, alas, remained. People say “use udev” on Linux. Clearly, they never have tried to cope with its crazy configuration. It’s a half-assed rule system. CLIPS? What? Naw, we’ll write our own and move the non-determinism to software. sigh.
    But here is how to handle it on Debian: http://debianclusters.org/index.php/Udev:_Renaming_Network_Interfaces
    Just don’t forget the sacrificial chicken.

  5. @Jason
    I don’t consider myself dumb (after my second cup of coffee that is) … and I have trouble parsing udev rules. The concept (rule based setup) is good. The implementation … well … to say it sucks is to be unfair … to things that merely suck. Udev is one of those very subtle beasts, which when it breaks, it breaks in such a way as to be nearly non-recoverable. We’ve had that happen before, in lab. There is almost nothing you can do to debug it.
    Yeah, I am getting pretty close to writing some code to fix this interactively. It annoys the customers. That isn’t a good thing.
    The nature of my code fix is gonna be something just like that (first live port is eth0) for the automatic mode, and the manual mode … well, I’ll try to make it drop dead easy.
    A customer spent 3 hours pulling their hair out over this. Ugh.

  6. speaking of udev-foo, I am working on a problem for a financial services customer this morning, and my udev mad-skillz aren’t quite where I need them. Running into udev just hanging now. Oh joy.

Comments are closed.