Python client for the OpenShift API
Branch: master
Clone or download
fabianvf Prevent API groups from growing stale (#264)
* Make api_groups a property

* Prevent api_groups from growing stale by re-requesting resources every time

* Add back v1.List

* Typo
Latest commit 31c96ef Feb 12, 2019

README.md

OpenShift python client

Build Status Coverage Status

Python client for the Kubernetes and OpenShift APIs.

There are two ways this project interacts with the Kubernets and OpenShift APIs. The first, now deprecated, is to use models and functions generated with swagger from the API spec. The second, new approach, is to use a single model and client to generically interact with all resources on the server. The dynamic client also works with resources that are defined by aggregated API servers or Custom Resource Definitions.

Table of Contents

Installation

From source:

git clone https://github.com/openshift/openshift-restclient-python.git
cd openshift-restclient-python
python setup.py install

From PyPi directly:

pip install openshift

Using Dockerfile:

docker build -t openshift-restclient-python -f Dockerfile .

Usage

The OpenShift client depends on the Kubernetes Python client, and as part of the installation process, the Kubernetes (K8s) client is automatically installed.

In the case you are using Docker, you will likely need to share your .kube/config with the openshift-restclient-python container:

docker run -it -v $HOME/.kube/config:/root/.kube/config:z openshift-restclient-python python

To work with the dynamic client, you will need an instantiated Kubernetes client object. The Kubernetes client object requires a Kubernetes Config that can be set in the Config class or using a helper utility. All of the examples that follow make use of the new_client_from_config() helper utility provided by the Kubernetes Client Config that returns an API client to be used with any API object. There are plenty of Kubernetes Client examples to examine other ways of accessing Kubernetes Clusters.

Examples

Create a Service

import yaml
from kubernetes import client, config
from openshift.dynamic import DynamicClient

k8s_client = config.new_client_from_config()
dyn_client = DynamicClient(k8s_client)

v1_services = dyn_client.resources.get(api_version='v1', kind='Service')

service = """
kind: Service
apiVersion: v1
metadata:
  name: my-service
spec:
  selector:
    app: MyApp
ports:
  - protocol: TCP
    port: 8080
    targetPort: 9376
"""

service_data = yaml.load(service)
resp = v1_services.create(body=service_data, namespace='default')

# resp is a ResourceInstance object
print(resp.metadata)

Create a Route

Now, we create a Route object, and associate it with the Service from our previous example:

import yaml
from kubernetes import client, config
from openshift.dynamic import DynamicClient

k8s_client = config.new_client_from_config()
dyn_client = DynamicClient(k8s_client)

v1_routes = dyn_client.resources.get(api_version='route.openshift.io/v1', kind='Route')

route = """
apiVersion: route.openshift.io/v1
kind: Route
metadata:
  name: frontend
spec:
  host: www.example.com
  to:
    kind: Service
    name: my-service
"""

route_data = yaml.load(route)
resp = v1_routes.create(body=route_data, namespace='default')

# resp is a ResourceInstance object
print(resp.metadata)

List Projects

The following uses the dynamic client to list Projects the user can access:

from kubernetes import client, config
from openshift.dynamic import DynamicClient

k8s_client = config.new_client_from_config()
dyn_client = DynamicClient(k8s_client)

v1_projects = dyn_client.resources.get(api_version='project.openshift.io/v1', kind='Project')

project_list = v1_projects.get()

for project in project_list.items:
    print(project.metadata.name)

Custom Resources

In the following example, we first create a Custom Resource Definition for foos.bar.com, then create an Foo resource, and finally get a list of Foo resources:

import yaml
from kubernetes import client, config
from openshift.dynamic import DynamicClient

k8s_client = config.new_client_from_config()
dyn_client = DynamicClient(k8s_client)

custom_resources = dyn_client.resources.get(
  api_version='apiextensions.k8s.io/v1beta1',
  kind='CustomResourceDefinition'
)

# Define the Foo Resource
foo_crd = """
kind: CustomResourceDefinition
apiVersion: apiextensions.k8s.io/v1beta1
metadata:
  name: foos.bar.com
spec:
  group: bar.com
  names:
    kind: Foo
    listKind: FooList
    plural: foos
    shortNames:
    - foo
    singular: foo
  scope: Namespaced
  version: v1beta1
"""
custom_resources.create(body=yaml.load(foo_crd))

foo_resources = None
while not foo_resources:
  try:
    # Notice the re-instantiation of the dynamic client as a new resource has been created.
    dyn_client = DynamicClient(k8s_client)
    foo_resources = dyn_client.resources.get(api_version='bar.com/v1beta1', kind='Foo')
  except:
    pass

# Create the Foo Resource
foo_resource_cr = """
kind: Foo
apiVersion: bar.com/v1beta1
metadata:
  name: example-foo
  namespace: default
spec:
  version: 1
"""
foo_resources.create(body=yaml.load(foo_resource_cr))

for item in foo_resources.get().items:
  print(item.metadata.name)

Available Methods for Resources

The generic Resource class supports the following methods, though every resource kind does not support every method.

Get

get(name=None, namespace=None, label_selector=None, field_selector=None, **kwargs)

Query for a resource in the cluster. Will return a ResourceInstance object or raise a NotFoundError

v1_services = dyn_client.resources.get(api_version='v1', kind='Service')

# Gets the specific Service named 'example' from the 'test' namespace
v1_services.get(name='example', namespace='test')

# Lists all Services in the 'test' namespace
v1_services.get(namespace='test')

# Lists all Services in the cluster (requires high permission level)
v1_services.get()

# Gets all Services in the 'test' namespace with the 'app' label set to 'foo'
v1_services.get(namespace='test', label_selector='app=foo')

# Gets all Services except for those in the 'default' namespace
v1_services.get(field_selector='metadata.namespace!=default')

get(body=None, namespace=None, **kwargs)

Query for a resource in the cluster. Will return a ResourceInstance object or raise a NotFoundError

For List kind resources (ie, the resource name ends in List), the get implementation is slightly different. Rather than taking a name, they take a *List kind definition and call get for each definition in the list.

v1_service_list = dyn_client.resources.get(api_version='v1', kind='ServiceList')

body = {
    'kind': 'ServiceList',
    'apiVersion': 'v1',
    'items': [
        'metadata': {'name': 'my-service'},
        'spec': {
            'selector': {'app': 'MyApp'},
            'ports': [{
                'protocol': 'TCP',
                'port': '8080',
                'targetPort': '9376'
            }]
        }
    ],
    # More definitions would go here
}
# Gets the specified Service(s) from the 'test' namespace
v1_service_list.get(body=body, namespace='test')

# Lists all Services in the 'test' namespace
v1_service_list.get(namespace='test')

# Lists all Services in the cluster (requires high permission level)
v1_service_list.get()

Create

create(body=None, namespace=None, **kwargs)

v1_services = dyn_client.resources.get(api_version='v1', kind='Service')

body = {
    'kind': 'Service',
    'apiVersion': 'v1',
    'metadata': {'name': 'my-service'},
    'spec': {
        'selector': {'app': 'MyApp'},
        'ports': [{
            'protocol': 'TCP',
            'port': '8080',
            'targetPort': '9376'
        }]
    }
}

# Creates the above service in the 'test' namespace
v1_services.create(body=body, namespace='test')

The create implementation is the same for *List kinds, except that each definition in the list will be created separately.

If the resource is namespaced (ie, not cluster-level), then one of namespace, label_selector, or field_selector is required.

If the resource is cluster-level, then one of name, label_selector, or field_selector is required.

Delete

delete(name=None, namespace=None, label_selector=None, field_selector=None, **kwargs)

v1_services = dyn_client.resources.get(api_version='v1', kind='Service')

# Deletes the specific Service named 'example' from the 'test' namespace
v1_services.delete(name='my-service', namespace='test')

# Deletes all Services in the 'test' namespace
v1_services.delete(namespace='test')

# Deletes all Services in the 'test' namespace with the 'app' label set to 'foo'
v1_services.delete(namespace='test', label_selector='app=foo')

# Deletes all Services except for those in the 'default' namespace
v1_services.delete(field_selector='metadata.namespace!=default')

delete(body=None, namespace=None, **kwargs)

For List kind resources (ie, the resource name ends in List), the delete implementation is slightly different. Rather than taking a name, they take a *List kind definition and call delete for each definition in the list.

v1_service_list = dyn_client.resources.get(api_version='v1', kind='ServiceList')

body = {
    'kind': 'ServiceList',
    'apiVersion': 'v1',
    'items': [
        'metadata': {'name': 'my-service'},
        'spec': {
            'selector': {'app': 'MyApp'},
            'ports': [{
                'protocol': 'TCP',
                'port': '8080',
                'tardeletePort': '9376'
            }]
        }
    ],
    # More definitions would go here
}
# deletes the specified Service(s) from the 'test' namespace
v1_service_list.delete(body=body, namespace='test')

# Deletes all Services in the 'test' namespace
v1_service_list.delete(namespace='test')

Patch

patch(body=None, namespace=None, **kwargs)

v1_services = dyn_client.resources.get(api_version='v1', kind='Service')

body = {
    'kind': 'Service',
    'apiVersion': 'v1',
    'metadata': {'name': 'my-service'},
    'spec': {
        'selector': {'app': 'MyApp2'},
    }
}

# patchs the above service in the 'test' namespace
v1_services.patch(body=body, namespace='test')

The patch implementation is the same for *List kinds, except that each definition in the list will be patched separately.

Replace

replace(body=None, namespace=None, **kwargs)

v1_services = dyn_client.resources.get(api_version='v1', kind='Service')

body = {
    'kind': 'Service',
    'apiVersion': 'v1',
    'metadata': {'name': 'my-service'},
    'spec': {
        'selector': {'app': 'MyApp2'},
        'ports': [{
            'protocol': 'TCP',
            'port': '8080',
            'targetPort': '9376'
        }]
    }
}

# replaces the above service in the 'test' namespace
v1_services.replace(body=body, namespace='test')

The replace implementation is the same for *List kinds, except that each definition in the list will be replaced separately.

Watch

watch(namespace=None, name=None, label_selector=None, field_selector=None, resource_version=None, timeout=None)

v1_services = dyn_client.resources.get(api_version='v1', kind='Service')

# Prints the resource that triggered each event related to Services in the 'test' namespace
for event in v1_services.watch(namespace='test'):
    print(event['object'])

DEPRECATED Generated client usage

To work with a K8s object, use the K8s client, and to work with an OpenShift specific object, use the OpenShift client. For example, the following uses the K8s client to create a new Service object:

import yaml
from kubernetes import client, config

config.load_kube_config()
api = client.CoreV1Api()

service = """
kind: Service
apiVersion: v1
metadata:
  name: my-service
spec:
  selector:
    app: MyApp
ports:
  - protocol: TCP
    port: 8080
    targetPort: 9376
"""

service_data = yaml.load(service)
resp = api.create_namespaced_service(body=service_data, namespace='default')

# resp is a V1Service object
print resp.metadata.self_link

Now in the following example, we use the OpenShift client to create a Route object, and associate it with the new Service:

import yaml
from openshift import client, config

config.load_kube_config()
api = client.OapiApi()

route = """
apiVersion: v1
kind: Route
metadata:
  name: frontend
spec:
  host: www.example.com
  to:
    kind: Service
    name: my-service
"""

route_data = yaml.load(route)
resp = api.create_namespaced_route(body=route_data, namespace='default')

# resp is a V1Route object
print resp.metadata.self_link

And finally, the following uses the OpenShift client to list Projects the user can access:

from openshift import client, config

config.load_kube_config()
oapi = client.OapiApi()

project_list = oapi.list_project()
for project in project_list.items:
    print project.metadata.name

DEPRECATED Documentation

All OpenShift API and Model documentation can be found in the Generated client's README file

DEPRECATED Update generated client

Updating the generated client requires the following tools:

  • tox
  • docker

To apply the updates:

  1. Incorporate new changes to update scripts
  2. Run tox -e update_client

Compatibility

We are downstream of the kubernetes python client. We maintain compatibility for API version n-2 - so if you are connecting to a version 3.6 OpenShift cluster, the list of supported python client versions would be [0.3.x, 0.4.x, 0.5.x].

Compatibility matrix

openshift python kubernetes python Kubernetes 1.5 Kubernetes 1.6 Kubernetes 1.7 Kubernetes 1.8 Kubernetes 1.9
openshift 0.3 kubernetes 3.0 + + - -
openshift 0.4 kubernetes 4.0 +* + + -
openshift 0.5 kubernetes 5.0 +* +* + +
openshift HEAD kubernetes HEAD +* +* + +

Key:

  • Exactly the same features / API objects in both openshift-restclient-python and the OpenShift version.
  • + openshift-restclient-python has features or api objects that may not be present in the OpenShift cluster, but everything they have in common will work.
  • - The OpenShift cluster has features the openshift-restclient-python library can't use (additional API objects, etc).
  • * This client/server combination may work, but is not officially supported.

Community, Support, Discussion

If you have any problem with the package or any suggestions, please file an issue.

Code of Conduct

Participation in the Kubernetes community is governed by the CNCF Code of Conduct.