tiny-kubernetes provides a thin wrapper on top of requests to make it easier to work with Kubernetes APIs.

• Automatic authentication ($KUBECONFIG)
• Simplified response traversal (never['do']['this']['again'])
• Sane exceptions
• No crazy complex client-side APIs to deal with

The official kubernetes package is, uh, complex. While it does support all the advanced features the Kubernetes API offers (like WebSocket streaming) it is often non-trivial to work with. tiny-kubernetes aims to provide most of the same functionality with minimal overhead.

To manually install, run:

pip install git+https://github.com/monasca/tiny-kubernetes.git

To depend on tiny-kubernetes from requirements.txt, add the following:

-e git+https://github.com/monasca/tiny-kubernetes.git#egg=tiny-kubernetes

(note: requires pip 7.0 or greater)

You can depend on a specific commit too:

-e git+https://github.com/monasca/tiny-kubernetes.git@[GIT SHA]#egg=tiny-kubernetes

from tiny_kubernetes import KubernetesAPIClient

client = KubernetesAPIClient()
client.load_auto_config()

pods = client.get('/api/v1/namespaces/{}/pods', 'default')
for pod in pods['items']:
  print pod.metadata.name

Notice that you can access (almost) all parts of the Kubernetes API response using dot accessors (courtesy of DotMap). As .items() is a reserved name in dictionaries, it needs to be accessed more traditionally, but most other member names can use the shortened syntax.

The endpoint is formatted using str.format with the remaining positional arguments. All keyword arguments are passed through to requests's Session.request, so you can include a body with json={...} or URL parameters with params={...}.

A special json_patch method exists to help perform patches. It works around a few potential requests bugs automatically. For example, it can be used to append labels to a resource:

client.json_patch([
  'op': 'add',
  'path': '/metadata/labels/foo',
  'value': 'bar'
], '/api/v1/namespaces/{}/pods/{}', 'default', 'some-pod')

• Temporary files may be created if using a local $KUBECONFIG with base64-encoded certificates. Due to limitations in python's SSL library these must be decoded and written to the filesystem to be used. They should be cleaned up when the interpreter exits.