Puppet.Daemon

Documentation

initDaemon :: Prefs -> IO (String -> Facts -> IO (Either String FinalCatalog))

This is a high level function, that will initialize the parsing and interpretation infrastructure from the Prefs structure, and will return a function that will take a node name, Facts and return either an error or the FinalCatalog.

It will internaly initialize several threads that communicate with channels. It should scale well, althrough it hasn't really been tested yet. It should cache the ASL of every .pp file, and could use a bit of memory. As a comparison, it fits in 60 MB with the author's manifests, but really breathes when given 300 MB of heap space. In this configuration, even if it spawns a ruby process for every template evaluation, it is way faster than the puppet stack.

It is recommended to ask for as many parser and interpreter threads as there are CPUs.

Known bugs :

• It might be buggy when top level statements that are not classdefinenodes are altered, or when files loaded with require are changed.
• Exported resources are not yet supported.
• The catalog is not computed exactly the same way Puppet does. Take a look at Puppet.Interpreter.Catalog for a list of differences.
• Parsing incompatibilities are listed in Puppet.DSL.Parser.
• There might be race conditions because file status are checked before they are opened. This means the program might end with an exception when the file is not existent. This will need fixing.