container-runtime-interface-api

container-runtime-interface-api

Python library for communication with the Kubernetes Container Runtime Interface API.

Usage

Install container-runtime-interface-api with pipenv or pip:

$ pipenv install container-runtime-interface-api
Adding container-runtime-interface-api to Pipfile's [packages]…
✔ Installation Succeeded
...

This project currently supports Python 3.10+.

Connection

Connection to the CRI API is generally done through a UNIX socket, but any gRPC address supported by insecure_channel will work.

from cri_api.channel import Channel
channel = Channel.from_env() # Loads from RUNTIME_SOCK
channel = Channel("unix:///var/run/dockershim.sock") # Explicit argument

Images

The Images class is a thin wrapper around the existing ImageService API:

from cri_api.images import Images

channel = Channel.from_env()
images = Images(channel)

images.list_images()
images.pull_image("busybox")

busybox_images = [i["id"] for i in images.list_images() if any("busybox" in r for r in i["repoTags"])]
[images.remove_image(i) for i in busybox_images]

Containers

The Containers class is a thin wrapper around the existing RuntimeService API:

from cri_api.images import Images
from cri_api import ContainerFilter, ContainerState, ContainerStateValue

channel = Channel.from_env()
images = Containers(channel)

containers.list_containers()
containers.list_containers(ContainerFilter(state=ContainerStateValue(state=ContainerState.CONTAINER_EXITED)))

containers.get_container("9d81052cc027a1fb2ec61b898ea0fd6fc88216ce730ad75f4c52b29849cb440f")

Raw

Raw access to the underlying CRI API objects can be done by importing from cri_api:

from os import getenv
from grpc import insecure_channel
from cri_api import RuntimeServiceStub, ListContainersRequest

stub = RuntimeServiceStub(insecure_channel(getenv("RUNTIME_SOCK")))
response = stub.ListContainers(ListContainersRequest())
containers = response.containers

Updating Protobuf Python Generated Code

$ bin/update-proto.sh

Commit & create a new pull request!

Development

Interactive development on MacOS can be done by leveraging minikube:

$ minikube start --container-runtime=cri-o
$ minikube ssh
$ socat -d -d TCP4-LISTEN:15432,fork UNIX-CONNECT:/var/run/crio/crio.sock

# In another window, you can now connect on $(minikube ip):15432
$ export RUNTIME_SOCK=$(minikube ip):15432
...

Testing

Unit tests are run against all supported Python versions: 3.10, 3.11, and 3.12.

A basic set of integration tests is included alongside the unit tests, but are only run if the RUNTIME_SOCK environment variable is set correctly. They are verified in CI against a minikube installation running Docker and dockershim.

Run unit tests with:

$ pytest