help-cfengine
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Help with cfengine architecture


From: Edward F. Brown
Subject: Re: Help with cfengine architecture
Date: Sun, 10 Dec 2006 12:28:56 -0700 (MST)
User-agent: SquirrelMail/1.4.8-2.el3.7lanl

On Sat, December 9, 2006 12:53 pm, Zeev Fisher wrote:
> I know that it might be possible to prepare some pre script which
> identify all relevant files in the tree and prepare every time input
> file for cfagent but any idea how to implement this without external
> script will be appreciated.

Maybe I misunderstand something, but I think you've answered your own
question above.  You don't need to separately compute needed input files,
and you don't need complicated excludes, when you've organized inputs into
a hierarchy.  You set some variables, and let the same copy action work
everywhere.  Or possibly a set of copy actions, to provide a range of
specifity, as Eric Sorenson describes here:
  http://cfwiki.org/cfwiki/index.php/Singlecopy_Nirvana

As for the actual structure of a hierarchy (and whether you use file per
service or file per class, or something else), I think there's some
arbitrariness involved.  No matter how you slice it, a diverse,
heterogenous environment is going to result in some complexity, possibly
even maintaining the same file in a couple of places, or resorting to
symlinks.  Probably other factors will shape the outcome more than mere
organizational logic.  E.g., do different groups of people need to
separately maintain parts of the configuration; is keeping Paris configs
out of the config files copied to London systems important...

What you propose looks ok as far as it goes, though it doesn't allow for
class-specific files: where to put a file for WebServers only?  Frankly, I
don't have a good answer.  I might add a class-specific directory where
needed, perhaps with a further breakdown, depending on how high in the
tree. It could be that there simply is no one organizational scheme that
would work everywhere, unless it became unmanageably complex.  Like LDAP,
there are some conventional 'lightweight' schemas in common use, and lots
of improvisations too.

-Ed




reply via email to

[Prev in Thread] Current Thread [Next in Thread]