Wild-eyed thinking aloud: Python System Management Infrastructure

[François Pinard]

[Ng Pheng Siong]

> According to William Annis  <annis at biostat.wisc.edu>:
> > Recently I have found myself designing a database-driven system
> > to keep track of our machines, what they do, where they sit, etc.,
> > and I keep thinking "William, you should write an *infrastructure*
> > for all this system junk you do."  I'd rather not do this alone.

> This problem has been around long before the first Internet gold rush -
> many skilled people have attacked it. Find something you can get going,
> e.g., cfengine, Big Brother, even the old venerable Nocol.

About `cfengine', I tried it very seriously and for a long while, and
despite seduced by the idea, came to regret having used this package, for
its lack of maintainability.  But as I became much dependent on `cfengine',
it was just not possible for me to dismiss it.  So, after having sufferred
enough, I bit the bullet, and rewrote those parts I needed in Python.
What a relief!  That was much, much more maintainable.  It was also easy
to use Python itself as a configuration description language, and this gave
me, for free, a _lot_ of flexibility.  Speed was also very acceptable: such
beasts are typically IO-bound, `cfengine' IO optimizations were not difficult
to transpose into Python.  (I'm not aware of Big Brother nor Nocol.)

> Join Usenix and SAGE.  Check out the material of Usenix's sysadmin
> conferences from years past.  Behold the diversity.  It doesn't have to
> be Python.

I quite understand your call for broader views.  Yet, in my own experience,
unless you are ready to fully accept a package for what it statically is,
"being Python" helps a great deal.

-- 
François Pinard   http://www.iro.umontreal.ca/~pinard