Kubernetes & OpenShift Java Client

This client provides access to the full Kubernetes &OpenShift REST APIs via a fluent DSL.

Contents

• Usage
  • Creating a client
  • Configuring the client
  • Loading resources from external sources
  • Passing a reference of a resource to the client
  • Adapting a client
    • Adapting and close
• Mocking Kubernetes
• Who Uses Fabric8 Kubernetes Client?
• Kubernetes Operators in Java Written using Fabric8 Kubernetes Client
• Kubernetes and Red Hat OpenShift Compatibility Matrix
• Kubernetes Client CHEAT SHEET
• Kubectl Java Equivalents
• FAQs

Usage

Creating a client

The easiest way to create a client is:

KubernetesClient client = new DefaultKubernetesClient();

DefaultOpenShiftClient implements both the KubernetesClient & OpenShiftClient interface so if you need theOpenShift extensions, such as Builds, etc then simply do:

OpenShiftClient osClient = new DefaultOpenShiftClient();

Configuring the client

This will use settings from different sources in the following order of priority:

• System properties
• Environment variables
• Kube config file
• Service account token & mounted CA certificate

System properties are preferred over environment variables. The following system properties & environment variables can be used for configuration:

Alternatively you can use the ConfigBuilder to create a config object for the Kubernetes client:

Config config = new ConfigBuilder().withMasterUrl("https://mymaster.com").build();
KubernetesClient client = new DefaultKubernetesClient(config);

Using the DSL is the same for all resources.

List resources:

NamespaceList myNs = client.namespaces().list();

ServiceList myServices = client.services().list();

ServiceList myNsServices = client.services().inNamespace("default").list();

Get a resource:

Namespace myns = client.namespaces().withName("myns").get();

Service myservice = client.services().inNamespace("default").withName("myservice").get();

Delete:

Namespace myns = client.namespaces().withName("myns").delete();

Service myservice = client.services().inNamespace("default").withName("myservice").delete();

Editing resources uses the inline builders from the Kubernetes Model:

Namespace myns = client.namespaces().withName("myns").edit(n -> new NamespaceBuilder(n)
                   .editMetadata()
                     .addToLabels("a", "label")
                   .endMetadata()
                   .build());

Service myservice = client.services().inNamespace("default").withName("myservice").edit(s -> new ServiceBuilder(s)
                     .editMetadata()
                       .addToLabels("another", "label")
                     .endMetadata()
                     .build());

In the same spirit you can inline builders to create:

Namespace myns = client.namespaces().create(new NamespaceBuilder()
                   .withNewMetadata()
                     .withName("myns")
                     .addToLabels("a", "label")
                   .endMetadata()
                   .build());

Service myservice = client.services().inNamespace("default").create(new ServiceBuilder()
                     .withNewMetadata()
                       .withName("myservice")
                       .addToLabels("another", "label")
                     .endMetadata()
                     .build());

You can also set the apiVersion of the resource like in the case of SecurityContextConstraints :

SecurityContextConstraints scc = new SecurityContextConstraintsBuilder()
		.withApiVersion("v1")
		.withNewMetadata().withName("scc").endMetadata()
		.withAllowPrivilegedContainer(true)
		.withNewRunAsUser()
		.withType("RunAsAny")
		.endRunAsUser()
		.build();

Following events

Use io.fabric8.kubernetes.api.model.Event as T for Watcher:

client.events().inAnyNamespace().watch(new Watcher<Event>() {

  @Override
  public void eventReceived(Action action, Event resource) {
    System.out.println("event " + action.name() + " " + resource.toString());
  }

  @Override
  public void onClose(KubernetesClientException cause) {
    System.out.println("Watcher close due to " + cause);
  }

});

Working with extensions

The kubernetes API defines a bunch of extensions like daemonSets, jobs, ingresses and so forth which are all usable in the extensions() DSL:

e.g. to list the jobs...

jobs = client.batch().jobs().list();

Loading resources from external sources

There are cases where you want to read a resource from an external source, rather than defining it using the clients DSL.For those cases the client allows you to load the resource from:

• A file (Supports both java.io.File and java.lang.String)
• A url
• An input stream

Once the resource is loaded, you can treat it as you would, had you created it yourself.

For example lets read a pod, from a yml file and work with it:

Pod refreshed = client.load('/path/to/a/pod.yml').fromServer().get();
Boolean deleted = client.load('/workspace/pod.yml').delete();
LogWatch handle = client.load('/workspace/pod.yml').watchLog(System.out);

Passing a reference of a resource to the client

In the same spirit you can use an object created externally (either a reference or using its string representation).

For example:

Pod pod = someThirdPartyCodeThatCreatesAPod();
Boolean deleted = client.resource(pod).delete();

Adapting the client

The client supports plug-able adapters. An example adapter is the OpenShift Adapterwhich allows adapting an existing KubernetesClient instance to an OpenShiftClient one.

For example:

KubernetesClient client = new DefaultKubernetesClient();

OpenShiftClient oClient = client.adapt(OpenShiftClient.class);

The client also support the isAdaptable() method which checks if the adaptation is possible and returns true if it does.

KubernetesClient client = new DefaultKubernetesClient();
if (client.isAdaptable(OpenShiftClient.class)) {
    OpenShiftClient oClient = client.adapt(OpenShiftClient.class);
} else {
    throw new Exception("Adapting to OpenShiftClient not support. Check if adapter is present, and that env provides /oapi root path.");
}

Adapting and close

Note that when using adapt() both the adaptee and the target will share the same resources (underlying http client, thread pools etc).This means that close() is not required to be used on every single instance created via adapt.Calling close() on any of the adapt() managed instances or the original instance, will properly clean up all the resources and thus none of the instances will be usable any longer.

Mocking Kubernetes

Along with the client this project also provides a kubernetes mock server that you can use for testing purposes.The mock server is based on https://github.com/square/okhttp/tree/master/mockwebserver but is empowered by the DSL and features provided by https://github.com/fabric8io/mockwebserver.

The Mock Web Server has two modes of operation:

• Expectations mode
• CRUD mode

Expectations mode

It's the typical mode where you first set which are the expected http requests and which should be the responses for each request.More details on usage can be found at: https://github.com/fabric8io/mockwebserver

This mode has been extensively used for testing the client itself. Make sure you check kubernetes-test.

To add a Kubernetes server to your test:

@Rule
public KubernetesServer server = new KubernetesServer();

CRUD mode

Defining every single request and response can become tiresome. Given that in most cases the mock webserver is used to perform simple crud based operations, a crud mode has been added.When using the crud mode, the mock web server will store, read, update and delete kubernetes resources using an in memory map and will appear as a real api server.

To add a Kubernetes Server in crud mode to your test:

@Rule
public KubernetesServer server = new KubernetesServer(true, true);

Then you can use the server like:

@Test
public void testInCrudMode() {
    KubernetesClient client = server.getClient();
    final CountDownLatch deleteLatch = new CountDownLatch(1);
    final CountDownLatch closeLatch = new CountDownLatch(1);

    //CREATE
    client.pods().inNamespace("ns1").create(new PodBuilder().withNewMetadata().withName("pod1").endMetadata().build());

    //READ
    podList = client.pods().inNamespace("ns1").list();
    assertNotNull(podList);
    assertEquals(1, podList.getItems().size());

    //WATCH
    Watch watch = client.pods().inNamespace("ns1").withName("pod1").watch(new Watcher<Pod>() {
        @Override
        public void eventReceived(Action action, Pod resource) {
            switch (action) {
                case DELETED:
                    deleteLatch.countDown();
                    break;
                default:
                    throw new AssertionFailedError(action.toString().concat(" isn't recognised."));
            }
        }

        @Override
        public void onClose(KubernetesClientException cause) {
            closeLatch.countDown();
        }
    });

    //DELETE
    client.pods().inNamespace("ns1").withName("pod1").delete();

    //READ AGAIN
    podList = client.pods().inNamespace("ns1").list();
    assertNotNull(podList);
    assertEquals(0, podList.getItems().size());

    assertTrue(deleteLatch.await(1, TimeUnit.MINUTES));
    watch.close();
    assertTrue(closeLatch.await(1, TimeUnit.MINUTES));
}

JUnit5 support through extension

You can use KubernetesClient mocking mechanism with JUnit5. Since it doesn't support @Rule and @ClassRule there is dedicated annotation @EnableKubernetesMockClient.If you would like to create instance of mocked KubernetesClient for each test (JUnit4 @Rule) you need to declare instance of KubernetesClient as shown below.

@EnableKubernetesMockClient
class ExampleTest {

    KubernetesClient client;

    @Test
    public void testInStandardMode() {
            ...
    }
}

In case you would like to define static instance of mocked server per all the test (JUnit4 @ClassRule) you need to declare instance of KubernetesClient as shown below.You can also enable crudMode by using annotation field crud.

@EnableKubernetesMockClient(crud = true)
class ExampleTest {

    static KubernetesClient client;

    @Test
    public void testInCrudMode() {
            ...
    }
}

Compatibility Matrix

Kubernetes Compatibility Matrix:

OpenShift Compatibility Matrix:

Note: This matrix is prepared by running our integration tests on different versions of OpenShift.

Major Changes in Kubernetes Client 4.0.0

All the resource objects used here will be according to OpenShift 3.9.0 and Kubernetes 1.9.0. All the resource objects will give all the fields according to OpenShift 3.9.0 and Kubernetes 1.9.0

• SecurityContextConstraints has been moved to OpenShift client from Kubernetes Client
• Job dsl is in both batch and extensions(Extensions is deprecated)
• DaemonSet dsl is in both apps and extensions(Extensions is deprecated)
• Deployment dsl is in both apps and extensions(Extensions is deprecated)
• ReplicaSet dsl is in both apps and extensions(Extensions is deprecated)
• NetworkPolicy dsl is in both network and extensions(Extensions is deprecated)
• Storage Class moved from client base DSL to storage DSL
• PodSecurityPolicies moved from client base DSL and extensions to only extensions
• ThirdPartyResource has been removed.

Who uses Kubernetes & OpenShift Java client?

Extensions:

• Istio API
• Service Catalog API
• Knative
• Tekton

Frameworks/Libraries/Tools:

• Arquillian Cube
• Apache Camel
• Apache Flink
• Apache Spark
• Jaeger Kubernetes
• Loom
• Microsoft Azure Libraries for Java
• Spinnaker Halyard
• Spring Cloud Connectors for Kubernetes
• Spring Cloud Kubernetes

CI Plugins:

• Deployment Pipeline Plugin (Jenkins)
• Kubernetes Eleastic Agent (GoCD)
• Kubernetes Plugin (Jenkins)
• Kubernetes Pipeline Plugin (Jenkins)
• OpenShift Sync Plugin (Jenkins)
• Kubernetes Plugin (Teamcity from Jetbrains)
• Kubernetes Agents for Bamboo (WindTunnel Technologies)

Build Tools:

• Fabric8 Maven Plugin
• Eclipse JKube
• Gradle Kubernetes Plugin

Platforms:

• Apache Openwhisk
• Eclipse che
• EnMasse
• Openshift.io (Launcher service)
• Spotify Styx
• Strimzi
• Syndesis
• StackGres
• Boomerang Flow

Proprietary Platforms:

• vCommander

As our community grows, we would like to track keep track of our users. Please send a PR with your organization/community name.

Tests we run for every new Pull Request

There are the links of the Github Actions and Jenkins for the tests which run for every new Pull Request. You can view all the recent builds also.

• Regression Tests
• Unit Tests

To get the updates about the releases, you can join https://groups.google.com/forum/embed/?place=forum/fabric8-devclients

Kubectl Java Equivalents

This table provides kubectl to Kubernetes Java Client mappings. Most of the mappings are quite straightforward and are one lineroperations. However, some might require slightly more code to achieve same result: