# Chef Metal Changelog

## 0.11.2 (6/4/2014)

- Fix issue where machines with different drivers could get default options from the global current driver

## 0.11.1 (6/4/2014)

- fix local mode port forwarding on IPv6 hosts

## 0.11 (6/4/2014)

- New Driver interface (see docs/ and blogs/ directories for documentation)
- New configuration (see docs/ and blogs/)
- get rid of annoying SSL warning (note: this turns off SSL verification, which was the default anyway)
- fix machine_batch error report to be less verbose
- fail when machine is being moved from driver to driver
- @marcusn disconnect from SSH when there is a problem
- fix SSH gateway code to honor any options given (@marcusn)
- Make machine_batch auto batching smarter (only batch things that have the same actions)
- Allow auto batching to be turned off with `auto_batch_machines = false` in recipes or config
- Allow this:
  ```ruby
  machine_batch do
    machine 'a'
    machine 'b'
  end
  ```
- Allow this:
  ```ruby
  machine_batch do
    machines 'a', 'b', 'c'
    action :destroy
  end
- fix issue setting Hosted Chef ACLs on nodes
- fix local mode forwarding in mixed IPv4/IPv6 environments

## 0.10.2 (5/2/2014)

- Fix crash with add_provisioner_options when provisioner_options is not yet set

## 0.10.1 (5/2/2014)

- Fix a crash when uploading files in a machine batch

## 0.10 (5/1/2014)

- Parallelism!
  - All machines by default will be created in parallel just before the first "machine" definition. They will attempt to run all the way to converge.  If they fail, add "with_machine_batch 'mybatch', :setup"
  - Use "with_machine_batch 'mybatch'" before any machines if you want tighter control. Actions include :delete, :acquire, :setup, and :converge.
- Parallelizableness: chef-metal now stores data in the run_context instead of globally, so that it can be run multiple times in parallel. This capability is not yet being used.

## 0.9.4 (4/23/2014)

- Preserve provisioner_output in machine resource (don't destroy it!!)

## 0.9.3 (4/13/2014)

- SSH: Treat EHOSTUNREACH as "machine not yet available" (helps with AWS)

## 0.9.2 (4/13/2014)

- Timeout stability fixes (makes EC2 a little stabler for some AMIs)

## 0.9.1 (4/11/2014)

- Make write_file and upload_file create parent directory

## 0.9 (4/9/2014)

- Add `files` and `file` attributes to the `machine` resource
- Fix `machine_execute` resource (@irvingpop)
- Fix `machine :converge` action (thanks @double-z)
- Make chef-client timeout longer by default (2 hours)
- Make chef_client_timeout a configurable option for all convergence strategies and provisioner_options
- Add `metal cp` command

## 0.8.2 (4/9/2014)

- Add timeout support to execute
- Fix machine_file resource
- Add ohai_hints DSL to machine resource (@xorl)

## 0.8.1 (4/9/2014)

- Bug: error! was not raising an error in the SSH and WinRM transports
- Transports: stream output automatically when in debug
- Support the :read_only execute hint (for Docker)
- Add more metal command lines (converge, update, delete)
- Add ChefMetal.connect_to_machine(machine_name) method to get Machine object for a node name

## 0.8 (4/8/2014)

- New machine_execute resource! (irving@getchef.com)
- Experimental "metal" command line: metal execute NODENAME COMMAND ARGS
- Transport: Add ability to stream execute() for better nested chef-client debugging

## 0.7 (4/5/2014)

- Change transport interface: add ability to rewrite URL instead of forwarding ports

## 0.6 (4/4/2014)

- Vagrant and Fog provisioners moved to their own gems (chef-metal-vagrant and chef-metal-fog)
- Support for Hosted and Enterprise Chef (https://github.com/dafyddcrosby)

## 0.5 (4/3/2014)

* Provisioner interface changes designed to allow provisioners to be used outside of Chef (doubt@getchef.com)
  * All Provisioner and Machine methods now take "action_handler" instead of "driver."  It uses the ActionHandler interface described in action_handler.rb.  In short:
    - driver.run_context -> action_handler.recipe_context
    - driver.updated_by_last_action(true) -> action_handler.updated!
    - driver.converge_by -> action_handler.perform_action
    - driver.cookbook_name -> driver.debug_name
  * Convergence strategy: delete_chef_objects() -> cleanup_convergence()
* Ability to get back to a machine from a node (another Provisioner interface change) (doubt@getchef.com):
  * Provisioners must create a file named `chef_metal/provisioner_init/<scheme>_init.rb`.  It will be required when a node is encountered with that scheme.  It should call ChefMetal.add_registered_provisioner_class(<scheme>, <provisioner class name>).  For the provisioner_url `fog:AWS:21348723432`, the scheme is "fog" and the file is `chef_metalprovisioner_init/fog_init.rb`.  It should call `ChefMetal.add_registered_provisioner_class('fog', ChefMetal::Provisioner::FogProvisioner)`.
  * Provisioner classes must implement the class method `inflate(node)`, which should create a Provisioner instance appropriate to the given `node` (generally by looking at `node['normal']['provisioner_output']`)
* New `NoConverge` convergence strategy that creates a node but does not install Chef or converge.
* Support for machine_file `group`, `owner` and `mode` attributes (@irvingpop)
* SSH transport (ryan@segv.net): try to enable pty when possible (increases chance of successful connection).  Set options[:ssh_pty_enable] to `false` to turn this off.  Set `true` to force it (and fail if we can't get it)

## 0.4 (3/29/2014)

* EC2: Make it possible for multiple IAM users to converge chef-metal on the same account
* Openstack: Openstack support via the Fog driver! (@cstewart87)
* EC2: Add :use_private_ip_for_ssh option, and use private ip by default if public IP does not exist.  (@xorl, @dafyddcrosby)
* RHEL/Centos: fix platform detection and installation