ctxswitch / stork Goto Github PK

Now that builder does not need to manage multiple tasks, it makes sense to merger the load class method and dsl delegators to the Collections object:

[Collections] Collections.load

Would parse the definition files and return the collections object

Add an 'allow' configuration option. By default the range will be an empty array which will allow all traffic. If values are given then the sinatra app will make a check in the before section. If the source IP is included then access will be allowed.

CIDR, dash seperated ranges, and single IP addresses are acceptable values.

Should be able to handle encrypted communication and minimal authentication.

When running 'host show' if the attribute has a nil value there is no newline.

Remove built in chef support in favor of creating tags or roles and build out some example post scripts to show how to integrate. This way we won't develop lock-ins surrounding tools or versions. Just make the core functionality flexible enough that it can be used with multiple tools.

Allow the additional '--kickstart' option for the HostShow plugin. If that option is given, dump the rendered kickstart template.

Before we start actually using this in a large production setting, get rid of webrick. It was a simple enough setup in the beginning, but as I look to deploy this into a larger production environment, I don't want to risk issues. Will also need to take into account reverse proxies that will potentially be in front of this as well.

Does the json option for chef-client handle setting default or override attributes? If so run_list should be changed to metadata and handle both run list items and attributes.

Only daemonize if specifically directed to on the command line.

Add a bundle init command to create initial bundle and intialize the repository

Set up key based authentication to the stork server that will allow service accounts and user accounts. This will be for updates only. I will assume that anyone using this for kickstarting will be using this in an isolated and safe environment so the retrieval of the kickstart files will be unauthenticated. This does leave us open for leaking passwords so best practice would be to always use a temporary password that will be changed during the configuration stage.

All changes to state will be audited and kept in system logs. They will include the time, account, and action.

Add a boot sync command plugin that regenerates all of the pxe boot configuration files.

Instead of setting a randomized password when midwife begins, generate a new one everytime the kickstart is rendered.

The bootstrap option should be set when generating the partition layout, a quick work around is to add it into the template.

Instead of building in full preseed support, make sure that we allow an option on the distro that will build kickstarts that can be used directly with ubuntu. Generated kickstart files will need to include a special preseed section that will contain some of the auto responses for questions that cannot be answered by kickstart.

Reference: https://help.ubuntu.com/community/KickstartCompatibility

Strip leading forward slash in paths before creating the url in the plugin fetch method

Need to add the ability to set STORK_CONFIG=/path/to/file in the environment so -c does not need to be passed in every time.

Allow a client to trigger a reload of all collections

Add validation for IP addresses. Make sure they are properly formed and nameservers, gateway... etc are within the subnet

NoMethodError - undefined method hashify' for nil:NilClass: /Users/rlyon/src/ruby/stork/lib/stork/server/application.rb:33:inblock in class:Application'

Split up the server and client configurations so we can keep them in different places as well as strip out the non essentials. The server config should be named server.rb and the client should be named client.rb

The server will still be found in /etc/stork by default but the client will be found in ~/.stork