>>>>> "Brandon" == Brandon S Allbery KF8NH <[EMAIL PROTECTED]> writes:

  Brandon> We support systems researchers; this leads to a real need
  Brandon> for one-off machines which nevertheless get our
  Brandon> configuration as much as possible, but with local
  Brandon> adjustments.  The "standard model" for configuration
  Brandon> management, on the other hand, assumes every machine in a
  Brandon> given class is effectively a clone.

We actually support a very similar environment. We are a DOE computer
science research facility of approximately 100 full time staff +
100-150 students and collaborators. Researchers are working on
everything from kernel development to system software development
(me) and high-level libraries and applications. 

We have a support model that includes three classes of systems:
 - Green (the systems group built it and maintains it)
 - Yellow (the systems group built it and maintains it, in conjunction
     with user(s))
 - Red (user built it, we know nothing)

We found that since moving to bcfg2, we've managed to switch almost
all of our red machines to yellow, and have even become agile enough
to move some yellow machines to green. So this model is pretty well
supported by bcfg2. Did I mention that my teeth are whiter too?

More seriously though, we have gotten a real handle on our
environment's configuration since deploying bcfg2, and have started to
be able to provide much better software services to our users. (which
directly led to the migration towards centrally supported systems)
lssconf-discuss mailing list

Reply via email to