docker_container: DeviceRequests.Capabilities are improperly validated

[skokec]

SUMMARY

When using device_requests with capabilities option for nvidia GPU in docker_container, ansible returns error indicating that capabilities are not formatted properly. This happens for the example provided in the documentation.

The issue seems to be around the line 1441 in docker_container.py where capabilitieslist is validated, however, this validation expects capabilities to be 'list of list of list of string' instead of 'list of list of string'. If I actually provide capabilities: [[[gpu]]] then this goes through but then I get docker marshal error indicating this is not what docker API expects.

A simple patch can fix this by removing the line 1441 in docker_container.py and updating the variable names. With this patch I do not get any issues and container is correctly created with the requested devices.

ISSUE TYPE

• Bug Report

COMPONENT NAME

docker_container

ANSIBLE VERSION

ansible 2.10.4
  config file = /etc/ansible/ansible.cfg
  configured module search path = [u'/home/domen/.ansible/plugins/modules', u'/usr/share/ansible/plugins/modules']
  ansible python module location = /usr/local/lib/python2.7/dist-packages/ansible
  executable location = /usr/bin/ansible
  python version = 2.7.17 (default, Sep 30 2020, 13:38:04) [GCC 7.5.0]

CONFIGURATION

OS / ENVIRONMENT

Linux ubuntu 4.15.0-118-generic #119-Ubuntu SMP Tue Sep 8 12:30:01 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux

STEPS TO REPRODUCE

- name: Start container with GPUs
  community.docker.docker_container:
    name: test
    image: ubuntu:18.04
    state: started
    device_requests:
      - # Add nVidia GPUs to this container
        driver: nvidia
        count: -1  # this means we want all
        capabilities:
          - - gpu
            - compute

EXPECTED RESULTS

Deploying container only on specific devices without errors

ACTUAL RESULTS

The following error:

Traceback (most recent call last):
  File \"/home/ubuntu/.ansible/tmp/ansible-tmp-1607364771.4-34439-220056238326582/AnsiballZ_docker_container.py\", line 102, in <module>
    _ansiballz_main()
  File \"/home/ubuntu/.ansible/tmp/ansible-tmp-1607364771.4-34439-220056238326582/AnsiballZ_docker_container.py\", line 94, in _ansiballz_main
    invoke_module(zipped_mod, temp_path, ANSIBALLZ_PARAMS)
  File \"/home/ubuntu/.ansible/tmp/ansible-tmp-1607364771.4-34439-220056238326582/AnsiballZ_docker_container.py\", line 40, in invoke_module
    runpy.run_module(mod_name='ansible_collections.community.docker.plugins.modules.docker_container', init_globals=None, run_name='__main__', alter_sys=True)
  File \"/usr/lib/python3.6/runpy.py\", line 205, in run_module
    return _run_module_code(code, init_globals, run_name, mod_spec)
  File \"/usr/lib/python3.6/runpy.py\", line 96, in _run_module_code
    mod_name, mod_spec, pkg_name, script_name)
  File \"/usr/lib/python3.6/runpy.py\", line 85, in _run_code
    exec(code, run_globals)
  File \"/tmp/ansible_community.docker.docker_container_payload_vj3muuqt/ansible_community.docker.docker_container_payload.zip/ansible_collections/community/docker/plugins/modules/docker_container.py\", line 3538, in <module>
  File \"/tmp/ansible_community.docker.docker_container_payload_vj3muuqt/ansible_community.docker.docker_container_payload.zip/ansible_collections/community/docker/plugins/modules/docker_container.py\", line 3529, in main
  File \"/tmp/ansible_community.docker.docker_container_payload_vj3muuqt/ansible_community.docker.docker_container_payload.zip/ansible_collections/community/docker/plugins/modules/docker_container.py\", line 2682, in __init__
  File \"/tmp/ansible_community.docker.docker_container_payload_vj3muuqt/ansible_community.docker.docker_container_payload.zip/ansible_collections/community/docker/plugins/modules/docker_container.py\", line 1447, in __init__
TypeError: 'str' object does not support item assignment

[felixfontein]

Thanks for reporting! I'll take a look at this later today or tomorrow. I would be glad if you could test a PR. (I don't have the means to properly test the option.)

[skokec]

Sure, I can test PR. Let me know when to do so.

[felixfontein]

@skokec please test it as soon as you can :)

[skokec]

It works OK now.

I've tested you branch fix-docker_container-device_requests with the following:

- hosts: <hostname>
  become: yes
  
  tasks:
    - name: Start container with GPUs
      community.docker.docker_container:
        name: test
        image: nvidia/cuda:10.1-runtime-ubuntu18.04
        state: started
        command: 'nvidia-smi -L'
        detach: false
        device_requests:
          - # Add nVidia GPUs to this container
            driver: nvidia
            device_ids: 
              - '0'
              - '1'
            capabilities:
              - ['gpu','nvidia']
      register: docker_container_output      
              
    - name: Show test output
      debug:
        msg: "{{ docker_container_output.container.Output  }}"

which correctly allocates first two GPUs:

TASK [Show test output] *******************************************************************************************
ok: [hostname] => {}

MSG:

GPU 0: GeForce RTX 2080 Ti (UUID: GPU-5185bf4e-d65e-9559-66aa-255ce2afeeb4)
GPU 1: GeForce RTX 2080 Ti (UUID: GPU-00ef98af-750e-2db7-ccbe-59054f6c8054)


PLAY RECAP ********************************************************************************************************
hostname                : ok=3    changed=1    unreachable=0    failed=0    skipped=0    rescued=0    ignored=0   

[felixfontein]

@skokec thanks! :)

[felixfontein]

I merged the PR and plan to release a new version of this collection later this week.

[felixfontein]

FYI, community.docker 1.0.1 has been released.