Asynchronous (AsyncIO) client library for the Kubernetes API.
This library is created in the same way as official https://github.com/kubernetes-client/python but uses asynchronous version of OpenAPI generator. My motivation is described here: kubernetes-client/python#324
From PyPi directly:
pip install kubernetes_asyncio
It requires Python 3.6+
To list all pods:
import asyncio
from kubernetes_asyncio import client, config
from kubernetes_asyncio.client.api_client import ApiClient
async def main():
# Configs can be set in Configuration class directly or using helper
# utility. If no argument provided, the config will be loaded from
# default location.
await config.load_kube_config()
# use the context manager to close http sessions automatically
async with ApiClient() as api:
v1 = client.CoreV1Api(api)
print("Listing pods with their IPs:")
ret = await v1.list_pod_for_all_namespaces()
for i in ret.items:
print(i.status.pod_ip, i.metadata.namespace, i.metadata.name)
if __name__ == '__main__':
loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.close()More complicated examples, like asynchronous multiple watch or tail logs from pods,
you can find in examples/ folder.
https://kubernetes-asyncio.readthedocs.io/
This library is generated in the same way as the official Kubernetes Python Library. It uses swagger-codegen and the same concepts like streaming, watching or reading configuration. Because of an early stage of this library some differences still exist:
| synchronous library kubernetes-client/python | this library | |
|---|---|---|
| authentication method | gcp-token, azure-token, user-token, oidc-token, user-password, in-cluster | gcp-token (only via gcloud command), user-token, oidc-token, user-password, in-cluster |
| streaming data via websocket from PODs | bidirectional | read-only is already implemented |
In case this library is used against Kubernetes cluster using client-go credentials plugin, the default asyncio event loop is SelectorEventLoop. This event loop selector, however, does NOT support pipes and subprocesses, so exec_provider.py::ExecProvider is failing. In order to avoid failures the ProactorEventLoop has to be selected. The ProactorEventLoop can be enabled via WindowsProactorEventLoopPolicy.
Application's code needs to contain following code:
import asyncio
asyncio.set_event_loop_policy(
asyncio.WindowsProactorEventLoopPolicy()
)This library is versioned in the same way as the synchronous library. The schema version has been changed with version v18.20.0. Now, first two numbers from version are Kubernetes version (v.1.18.20). The last number is for changes in the library not directly connected with K8s.
Install the development packages:
pip install -r requirements.txt
pip install -r test-requirements.txtYou can run the style checks and tests with
flake8 kubernetes_asyncio/
isort --diff kubernetes_asyncio/
py.test
