2011/3/10 David Lee <david....@ecmwf.int>: > Sorry, Mark. Yes, there was an element of letting off steam, partly > because I'm still struggling, despite significant cf-2 experience, to > get my poor little head around cf-3. (I could really, really do with a > "cookbook"; and with step-by-step definitions and examples. The > definitions and descriptions of cf-3 concepts such as bundles and bodies > and components etc. seem to be incomplete or fuzzy. (And half the > battle is that I don't even know how frame the questions that I should > be asking, so I don't know even what to ask.)
Well, in general "body" and "bundle" are just two different types of parsers. They are documented in short here: http://www.cfengine.org/manuals/cf3-reference.html#Containers The generalized bundle syntax is: bundle bundle_type bundle_name { section_name: class_expression:: "promise" option_name => option_value, option_name => option_value; } Here bundle_type can be: - server -- there's only one such bundle, its bundle_name is access_rules(), and it provides permissions setup for cf-serverd. - common -- a special kind of bundle which doesn't need to be included into the bundlesequence 'cause common bundles are read before bundlesequence starts. Common bundles are read on the server side, too, so they can be used to declare global variables and classes both for cf-agent and for cf-serverd. - agent -- any bundle included into bundlesequence directly (via body common control) or indirectly (via usebundle in methods). - edit_line -- a special kind of bundle used to provide bundle parser for file editing promises. The following section names apply for bundle agent XXXX: vars classes packages methods files processes commands reports The following section names apply for bundle edit_line XXXX: vars classes delete_lines insert_lines field_edits replace_patterns The generalized body syntax is: body body_type body_name { class_expression:: option_name => option_value; } Bodies are used to hide complex details of promises options, except for control bodies which define execution flow for Cfengine parts. One special body defines the whole processing flow for cf-agent: body common control { bundlesequence => { ... }; } Option names and values depend on the relevant section and/or body type. If a bundle or a body is called by a promise, the option_name refers to the bundle or body type (e.g. edit_line or package_method) and option_value contains its name: packages: "foo" package_method => rpm; Hope, it helps. > > The arrival of the "services" concept in cf-3 sounds absolutely right > and good, and makes better sense, as a higher-level abstraction, in many > cases than dabbling the a service's underlying processes. But then the > documentation leads me to believe that "services" is not available for > me in the current community edition. > > (I'm familiar with some OSS products having commercial siblings, for > instance CentOS/Redhat sendmail.org/sendmail.com. And I'm happy with > that, usually because the commercial one is adding support; and where it > is adding actual functionality, that tends to be relatively at the edges > rather than in the core. In "cfengine" it struck me that this > potentially excellent "services" concept felt core (hence, OSS). > especially as it is simply another subpoint in simply another subsection > of the documentation: Ref Manual 7.15.) > > Anyway, I'll see what I can do with processes. And if there is an RPM > of the new "process_stop" under "-n" fix, I'll test that. And I'll try > to come up with some well-framed questions, and positive suggestions for > documentation. > > Thanks. > > -- > David Lee > > > Mark Burgess wrote: >> David, >> >> if by open source you mean "free beer" then, your worst fears are >> confirmed. There is a paid version of cfengine, and it is better than >> the free one! (The audacity!) However, it is still apparently free to >> let off steam, which seems to be your main aim here. >> >> In fact services were introduced for the Windows version, and have not >> yet been implemented on Unix at all (since one has a perfectly good >> processes abstraction). The error message as such is misleading, since >> it is the Windows version of Cfengine which is Nova only. I expect you >> find that more acceptable. >> >> Mark > _______________________________________________ > Help-cfengine mailing list > Help-cfengine@cfengine.org > https://cfengine.org/mailman/listinfo/help-cfengine > -- SY, Seva Gluschenko. _______________________________________________ Help-cfengine mailing list Help-cfengine@cfengine.org https://cfengine.org/mailman/listinfo/help-cfengine
