Destroying images

[matiasdecarli]

Rigth now there is a bug when trying to destroy images.
Here is the problem

Chef::Log.debug("Destroying image: chef:#{container_name}")
image = Docker::Image.get("chef:#{container_name}")
image.delete

But in docker-api docs says that we should get the image by his ID

# Get Image from the server, with id
Docker::Image.get('df4f1bdecf40')

[matiasdecarli]

Ok, I actually solved this using

target_repository = 'chef'
image = find_image(target_repository, container_name)

if image == nil
  Chef::Log.debug("Image not found!")
else
  Chef::Log.debug("Removing image!")        
  image.delete
end

But im getting an error here container.delete when the container is running, even if right before theres a container.stop

If the container is stopped, this delete the container and the first part, deletes the images. Any idea why?

This is the log

[2014-10-07T15:14:04+00:00] DEBUG: Destroying container: deploy-8223
[2014-10-07T15:14:04+00:00] DEBUG: Stopping deploy-8223
[2014-10-07T15:14:04+00:00] DEBUG: Removing deploy-8223


    ================================================================================
    Error executing action `destroy` on resource 'machine[deploy-8223]'
    ================================================================================

    Docker::Error::ServerError
    --------------------------
    Expected(200..204) <=> Actual(500 InternalServerError)

    Resource Declaration:
    ---------------------
    # In /home/vagrant/cookbooks/matiasdecarli/recipes/create_container.rb

      7: machine 'deploy-8223'  do  
      8:   action :destroy
      9: end
     10: 

    Compiled Resource:
    ------------------
    # Declared in /home/vagrant/cookbooks/matiasdecarli/recipes/create_container.rb:7:in `from_file'

    machine("deploy-8223") do
      action [:destroy]
      retries 0
      retry_delay 2
      guard_interpreter :default
      chef_server {:chef_server_url=>"https://api.opscode.com/organizations/matiasdecarli", :options=>{:client_name=>"matiasdecarli", :signing_key_filename=>"/etc/chef/client.pem"}}
      driver "docker"
      cookbook_name :matiasdecarli
      recipe_name "create_container"
    end

[2014-10-07T15:14:16+00:00] INFO: Running queued delayed notifications before re-raising exception
[2014-10-07T15:14:16+00:00] DEBUG: Re-raising exception: Docker::Error::ServerError - machine[deploy-8223] (matiasdecarli::create_container line 7) had an error: Docker::Error::ServerError: Expected(200..204) <=> Actual(500 InternalServerError)

/var/lib/gems/1.9.1/gems/docker-api-1.13.0/lib/docker/connection.rb:54:in `rescue in request'
  /var/lib/gems/1.9.1/gems/docker-api-1.13.0/lib/docker/connection.rb:37:in `request'
  /var/lib/gems/1.9.1/gems/docker-api-1.13.0/lib/docker/connection.rb:61:in `block (2 levels) in <class:Connection>'
  /var/lib/gems/1.9.1/gems/docker-api-1.13.0/lib/docker/container.rb:134:in `remove'
  /var/lib/gems/1.9.1/gems/chef-metal-docker-0.4.3/lib/chef_metal_docker/docker_driver.rb:163:in `destroy_machine'

Update: I've solved this catching the exception. Probably not the best way to do it, but it does the trick

[matiasdecarli]

I can make a pull request about this @jkeiser

[jkeiser]

Sounds great!

[matiasdecarli]

Converted into PR

[jkeiser]

Reading the docs, this is hiding a 500 (meaning there is an issue with the Docker code). Is there a corresponding bug filed in Docker itself?

I'd feel a lot more comfortable here if we did an if container.exists? (or its equivalent) rather than catching all 500's and saying they are OK.

[randomcamel]

Do we really want to destroy any containers when we destroy the image? The Docker CLI tool requires -f to delete images used by live containers.