Operator configuration examples

Reference image streams (.spec.applicationImage)

To deploy an image from an image stream, you must specify a .spec.applicationImage field in your CR.

spec:
  applicationImage: my-namespace/my-image-stream:1.0

The previous example looks up the 1.0 tag from the my-image-stream image stream in the my-namespace project and populates the CR .status.imageReference field with a referenced image such as image-registry.openshift-image-registry.svc:5000/my-namespace/my-image-stream@sha256:*****. The operator watches the specified image stream and deploys new images as new ones are available for the specified tag.

To reference an image stream, the .spec.applicationImage field must follow the project_name/image_stream_name[:tag] format. If project_name or tag is not specified, the operator uses the default values of the CR namespace and of latest. For example, the applicationImage: my-image-stream configuration is the same as the applicationImage: my-namespace/my-image-stream:latest configuration.

The operator tries to find an image stream name first with the project_name/image_stream_name format and falls back to the registry lookup if it can't to find any image stream that matches the value.

This feature is only available if you are running on Red Hat® OpenShift®. The operator requires ClusterRole permissions if the image stream resource is in another namespace.

Configure service account (.spec.serviceAccount)

The operator can create a ServiceAccount resource when it deploys a WebSphereLibertyApplication CR. If .spec.serviceAccount.name is not specified in a CR, the operator creates a service account with the same name as the CR, such as my-app.

You can set .spec.serviceAccount.mountToken to disable mounting the service account token into the application pods. By default, the service account token is mounted. This configuration applies to either the default service account that the operator creates or to the custom service account that you provide.

If applications require specific permissions but still want the operator to create a ServiceAccount, you can manually create a role binding to bind a role to the service account that the operator created. To learn more about role-based access control (RBAC), see the Kubernetes documentation.

Add or change labels (.metadata.labels)

By default, the operator adds the following labels into all resources that are created for a WebSphereLibertyApplication CR.

You can add new labels or overwrite existing labels, excluding the app.kubernetes.io/instance label. To set labels, specify them in your CR as key-value pairs in the .metadata.labels field.

apiVersion: liberty.websphere.ibm.com/v1
Kind: WebSphereLibertyApplication
metadata:
  name: my-app
  labels:
    my-label-key: my-label-value
spec:
  applicationImage: quay.io/my-repo/my-app:1.0

After the initial deployment of the CR, any changes to its labels are applied only if a spec field is updated.

When running in Red Hat OpenShift, there are additional labels and annotations that are standard on the platform. Overwrite defaults where applicable and add any labels from the Red Hat OpenShift list that are not set by default using the previous instructions.

Add annotations (.metadata.annotations)

To add new annotations into all resources created for a WebSphere Liberty operator, specify them in your CR as key-value pairs in the .metadata.annotations field. Annotations in a CR override any annotations specified on a resource, except for the annotations set on Service with .spec.service.annotations.

apiVersion: liberty.websphere.ibm.com/v1
Kind: WebSphereLibertyApplication
metadata:
  name: my-app
  annotations:
    my-annotation-key: my-annotation-value
spec:
  applicationImage: quay.io/my-repo/my-app:1.0

After the initial deployment of the CR, any changes to its annotations are applied only if a spec field is updated.

When running in Red Hat OpenShift, there are additional annotations that are standard on the platform. Overwrite defaults where applicable and add any labels from the Red Hat OpenShift list that are not set by default using the previous instructions.

Set environment variables for an application container (.spec.env or .spec.envFrom)

To set environment variables for your application container, specify .spec.env or .spec.envFrom fields in a CR. The environment variables can come directly from key-value pairs, ConfigMap, or Secret. The environment variables set by the .spec.env or .spec.envFrom fields override any environment variables that are specified in the container image.

Use .spec.envFrom to define all data in a ConfigMap or a Secret as environment variables in a container. Keys from ConfigMap or Secret resources become environment variable names in your container. The following CR sets key-value pairs in .spec.env and .spec.envFrom fields.

spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  env:
    - name: DB_NAME
      value: "database"
    - name: DB_PORT
      valueFrom:
        configMapKeyRef:
          name: db-config
          key: db-port
    - name: DB_USERNAME
      valueFrom:
        secretKeyRef:
          name: db-credential
          key: adminUsername
    - name: DB_PASSWORD
      valueFrom:
        secretKeyRef:
          name: db-credential
          key: adminPassword
  envFrom:
    - configMapRef:
        name: env-configmap
    - secretRef:
        name: env-secrets

For another example that uses .spec.envFrom.secretRef, see Using environment variables for basic authentication credentials. For an example that overrides the console logging environment variable default values, see Override console logging environment variable default values (.spec.env).

Override console logging environment variable default values (.spec.env)

The WebSphere Liberty operator sets environment variables that are related to console logging by default. You can override the console logging default values with your own values in your CR .spec.env list.

The following table lists the console logging environment variables and their default values.

To override default values for the console logging environment variables, set your preferred values manually in your CR .spec.env list. For information about values that you can set, see the Open Liberty logging documentation.

The following example shows a CR .spec.env list that sets nondefault values for the console logging environment variables.

spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  env:
    - name: WLP_LOGGING_CONSOLE_FORMAT
      value: "DEV"
    - name: WLP_LOGGING_CONSOLE_SOURCE
      value: "messages,trace,accessLog"
    - name: WLP_LOGGING_CONSOLE_LOGLEVEL
      value: "error"

For more information about overriding variable default values, see Set environment variables for an application container (.spec.env or .spec.envFrom). For information about monitoring applications and analyzing application logs, see Observing with the WebSphere Liberty operator

Configure multiple application instances for high availability (.spec.replicas or .spec.autoscaling)

To run multiple instances of your application for high availability, use the .spec.replicas field for multiple static instances or the .spec.autoscaling field for auto-scaling, which autonomically creates or deletes instances based on resource consumption. The .spec.autoscaling.maxReplicas and .spec.resources.requests.cpu fields are required for auto-scaling.

Set privileges and permissions for a pod or container (.spec.securityContext)

A security context controls privilege and permission settings for a pod or application container. By default, the operator sets several .spec.securityContext parameters for an application container as shown in the following example.

spec:
  containers:
    - name: app
      securityContext:
        capabilities:
          drop:
            - ALL
        privileged: false
        runAsNonRoot: true
        readOnlyRootFilesystem: false
        allowPrivilegeEscalation: false
        seccompProfile:
          type: RuntimeDefault

To override the default values or set more parameters, change the .spec.securityContext parameters, for example:

spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  securityContext:
    readOnlyRootFilesystem: true
    runAsUser: 1001
    seLinuxOptions:
      level: "s0:c123,c456"

For more information, see Set the security context for a Container. For more information about security context parameters, see SecurityContext v1 core

Persist resources (.spec.statefulSet and .spec.volumeMounts)

If storage is specified in the WebSphereLibertyApplication CR, the operator can create a StatefulSet and PersistentVolumeClaim for each pod. If storage is not specified, StatefulSet resource is created without persistent storage.

The following CR definition uses .spec.statefulSet.storage to provide basic storage. The operator creates a StatefulSet with the size of 1Gi that mounts to the /data folder.

apiVersion: liberty.websphere.ibm.com/v1
Kind: WebSphereLibertyApplication
metadata:
  name: my-app
spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  statefulSet:
    storage:
      size: 1Gi
      mountPath: "/data"

A WebSphereLibertyApplication CR definition can provide more advanced storage. With the following CR definition, the operator creates a PersistentVolumeClaim called pvc with the size of 1Gi and ReadWriteOnce access mode. The operator enables users to provide an entire .spec.statefulSet.storage.volumeClaimTemplate for full control over the automatically created PersistentVolumeClaim. To persist to more than one folder, the CR definition uses the .spec.volumeMounts field instead of .spec.statefulSet.storage.mountPath.

apiVersion: liberty.websphere.ibm.com/v1
Kind: WebSphereLibertyApplication
metadata:
  name: my-app
spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  volumeMounts:
  - name: pvc
    mountPath: /data_1
    subPath: data_1
  - name: pvc
    mountPath: /data_2
    subPath: data_2
  statefulSet:
    storage:
      volumeClaimTemplate:
        metadata:
          name: pvc
        spec:
          accessModes:
          - "ReadWriteMany"
          storageClassName: 'glusterfs'
          resources:
            requests:
              storage: 1Gi

The following CR definition does not specify storage and creates StatefulSet resources without persistent storage. You can create StatefulSet resources without storage if you require only ordering and uniqueness of a set of pods.

apiVersion: liberty.websphere.ibm.com/v1
Kind: WebSphereLibertyApplication
metadata:
  name: my-app
spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  statefulSet: {}

Monitor resources (.spec.monitoring)

A WebSphere Liberty operator can create a ServiceMonitor resource to integrate with Prometheus Operator.

At minimum, provide a label for Prometheus set on ServiceMonitor objects. In the following example, the .spec.monitoring label is apps-prometheus.

spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  monitoring:
    labels:
       app-prometheus: ''
    endpoints:
    - interval: '30s'
      basicAuth:
        username:
          key: username
          name: metrics-secret
        password:
          key: password
          name: metrics-secret
      tlsConfig:
        insecureSkipVerify: true

For more advanced monitoring, set many ServicerMonitor parameters such as authentication secret with Prometheus Endpoint.

spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  monitoring:
    labels:
       app-prometheus: ''
    endpoints:
    - interval: '30s'
      basicAuth:
        username:
          key: username
          name: metrics-secret
        password:
          key: password
          name: metrics-secret
      tlsConfig:
        insecureSkipVerify: true

Specify multiple service ports (.spec.service.port* and .spec.monitoring.endpoints)

To provide multiple service ports in addition to the primary service port, configure the primary service port with the .spec.service.port, .spec.service.targetPort, .spec.service.portName, and .spec.service.nodePort fields. The primary port is exposed from the container that runs the application and the port values are used to configure the Route (or Ingress), Service binding and Knative service.

To specify an alternative port for Service Monitor, use the .spec.monitoring.endpoints field and specify either the port or targetPort field, otherwise it defaults to the primary port.

Specify the primary port with the .spec.service.port field and additional ports with the .spec.service.ports field as shown in the following example.

spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  service:
    type: NodePort
    port: 9080
    portName: http
    targetPort: 9080
    nodePort: 30008
    ports:
      - port: 9443
        name: https
  monitoring:
    endpoints:
      - basicAuth:
          password:
            key: password
            name: metrics-secret
          username:
            key: username
            name: metrics-secret
        interval: 5s
        port: https
        scheme: HTTPS
        tlsConfig:
          insecureSkipVerify: true
    labels:
      app-monitoring: 'true'

Configure probes (.spec.probes)

Probes are health checks on an application container to determine whether it is alive or ready to receive traffic. The WebSphere Liberty operator has startup, liveness, and readiness probes.

Probes are not enabled in applications by default. To enable a probe with the default values, set the probe parameters to {}. The following example enables all 3 probes to use default values.

apiVersion: liberty.websphere.ibm.com/v1
Kind: WebSphereLibertyApplication
metadata:
  name: my-app
spec:
  probes:
    startup: {}
    liveness: {}
    readiness: {}

  httpGet:
    path: /health/started
    port: 9443
    scheme: HTTPS
  timeoutSeconds: 2
  periodSeconds: 10
  failureThreshold: 20

  httpGet:
    path: /health/live
    port: 9443
    scheme: HTTPS
  initialDelaySeconds: 60
  timeoutSeconds: 2
  periodSeconds: 10
  failureThreshold: 3

  httpGet:
    path: /health/ready
    port: 9443
    scheme: HTTPS
  initialDelaySeconds: 10
  timeoutSeconds: 2
  periodSeconds: 10
  failureThreshold: 10

To override a default value, specify a different value. The following example overrides a liveness probe initial delay default of 60 seconds and sets the initial delay to 90 seconds.

spec:
  probes:
    liveness:
      initialDelaySeconds: 90

When a probe initialDelaySeconds parameter is set to 0, the default value is used. To set a probe initial delay to 0, define the probe instead of using the default probe. The following example overrides the default value and sets the initial delay to 0.

spec:
  probes:
    liveness:
      httpGet:
        path: "/health/live"
        port: 9443
      initialDelaySeconds: 0

Deploy serverless applications with Knative (.spec.createKnativeService)

If Knative is installed on a Kubernetes cluster, to deploy serverless applications with Knative on the cluster, the operator creates a Knative Service resource which manages the entire life cycle of a workload. To create a Knative service, set .spec.createKnativeService to true.

spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  createKnativeService: true

The operator creates a Knative service in the cluster and populates the resource with applicable WebSphereLibertyApplication fields. Also, it ensures non-Knative resources such as Kubernetes Service, Route, and Deployment are deleted.

The CRD fields that can populate the Knative service resource include .spec.applicationImage, .spec.serviceAccount.name, .spec.probes.liveness, .spec.probes.readiness, .spec.service.Port, .spec.volumes, .spec.volumeMounts, .spec.env, .spec.envFrom, .spec.pullSecret and .spec.pullPolicy. Startup probe is not fully supported by Knative, thus .spec.probes.startup does not apply when Knative service is enabled.

For details on how to configure Knative for tasks such as enabling HTTPS connections and setting up a custom domain, see the Knative documentation.

Autoscaling fields in WebSphereLibertyApplication are not used to configure Knative Pod Autoscaler (KPA). To learn how to configure KPA, see Configuring the Autoscaler.

Expose applications externally (.spec.expose, .spec.createKnativeService, .spec.route)

Expose an application externally with a Route, Knative Route, or Ingress resource.

To expose an application externally with a route in a non-Knative deployment, set .spec.expose to true.

The operator creates a secured route based on the application service when .spec.manageTLS is enabled. To use custom certificates, see information about .spec.service.certificateSecretRef and .spec.route.certificateSecretRef.

apiVersion: liberty.websphere.ibm.com/v1
Kind: WebSphereLibertyApplication
metadata:
  name: my-app
spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  expose: true

To expose an application externally with Ingress in a non-Knative deployment, complete the following steps.

1. To use the Ingress resource to expose your cluster, install an Ingress controller such a Nginx or Traefik.
2. Ensure that a Route resource is not on the cluster. The Ingress resource is created only if the Route resource is not available on the cluster.
3. To use the Ingress resource, set the defaultHostName variable in the Operator ConfigMap object to a hostname such as mycompany.com
4. Enable TLS. Generate a certificate and specify the secret that contains the certificate with the .spec.route.certificateSecretRef field.

   apiVersion: liberty.websphere.ibm.com/v1
   Kind: WebSphereLibertyApplication
   metadata:
     name: my-app
     namespace: backend
   spec:
     applicationImage: quay.io/my-repo/my-app:1.0
     expose: true
     route:
       certificateSecretRef: mycompany-tls

5. Specify .spec.route.annotations to configure the Ingress resource. Annotations such as Nginx, HAProxy, Traefik, and others are specific to the Ingress controller implementation.

   The following example specifies annotations, an existing TLS secret, and a custom hostname.

   apiVersion: liberty.websphere.ibm.com/v1
   Kind: WebSphereLibertyApplication
   metadata:
     name: my-app
     namespace: backend
   spec:
     applicationImage: quay.io/my-repo/my-app:1.0
     expose: true
     route:
       annotations:
         # You can use this annotation to specify the name of the ingress controller to use.
         # You can install multiple ingress controllers to address different types of incoming traffic such as an external or internal DNS.
         kubernetes.io/ingress.class: "nginx"
   
         # The following nginx annotation enables a secure pod connection:
         nginx.ingress.kubernetes.io/ssl-redirect: true
         nginx.ingress.kubernetes.io/backend-protocol: "HTTPS"
   
         # The following traefik annotation enables a secure pod connection:
         traefik.ingress.kubernetes.io/service.serversscheme: https
   
       # Use a custom hostname for the Ingress
       host: app-v1.mycompany.com
       # Reference a pre-existing TLS secret:
       certificateSecretRef: mycompany-tls

To expose an application as a Knative service, set .spec.createKnativeService and .spec.expose to true. The operator creates an unsecured Knative route. To configure secure HTTPS connections for your Knative deployment, see Configuring HTTPS with TLS certificates.

apiVersion: liberty.websphere.ibm.com/v1
Kind: WebSphereLibertyApplication
metadata:
  name: my-app
spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  createKnativeService: true
  expose: true

Allowing or limiting incoming traffic (.spec.networkPolicy)

By default, network policies for an application isolate incoming traffic.

• The default network policy created for applications that are not exposed limits incoming traffic to pods in the same namespace that are part of the same application. Traffic is limited to only the ports that are configured by the service. By default, traffic will be exposed to .spec.service.targetPort when specified and otherwise fallback to using the .spec.service.port. Using the same logic, traffic will be exposed for each additional targetPort or port provided in the .spec.service.ports[] array.
• Red Hat OpenShift supports network policies by default. For exposed applications on Red Hat OpenShift, the network policy allows incoming traffic from the Red Hat OpenShift ingress controller on the ports in the service configuration. The network policy also allows incoming traffic from the Red Hat OpenShift monitoring stack.
• For exposed applications on other Kubernetes platforms, the network policy allows incoming traffic from any pods in any namespace on the ports in the service configuration. For deployments to other Kubernetes platforms, ensure that your network plug-in supports the Kubernetes network policies.

To disable the creation of network policies for an application, set .spec.networkPolicy.disable to true.

spec:
  networkPolicy:
    disable: true

You can change the network policy to allow incoming traffic from specific namespaces or pods. By default, .spec.networkPolicy.namespaceLabels is set to the same namespace to which the application is deployed, and .spec.networkPolicy.fromLabels is set to pods that belong to the same application specified by .spec.applicationName. The following example allows incoming traffic from pods that are labeled with the frontend role and are in the same namespace.

spec:
  networkPolicy:
    fromLabels:
      role: frontend

The following example allows incoming traffic from pods that belong to the same application in the example namespace.

spec:
  networkPolicy:
    namespaceLabels:
      kubernetes.io/metadata.name: example

The following example allows incoming traffic from pods that are labeled with the frontend role in the example namespace.

spec:
  networkPolicy:
    namespaceLabels:
      kubernetes.io/metadata.name: example
    fromLabels:
      role: frontend

Bind applications with operator-managed backing services (.status.binding.name and .spec.service.bindable)

The Service Binding Operator enables application developers to bind applications together with operator-managed backing services. If the Service Binding Operator is installed on your cluster, you can bind applications by creating a ServiceBindingRequest custom resource.

You can configure a WebSphere Liberty application to behave as a Provisioned Service that is defined by the Service Binding Specification. According to the specification, a Provisioned Service resource must define a .status.binding.name that refers to a Secret. To expose your application as a Provisioned Service, set the .spec.service.bindable field to a value of true. The operator creates a binding secret that is named CR_NAME-expose-binding and adds the host, port, protocol, basePath, and uri entries to the secret.

To override the default values for the entries in the binding secret or to add new entries to the secret, create an override secret that is named CR_NAME-expose-binding-override and add any entries to the secret. The operator reads the content of the override secret and overrides the default values in the binding secret.

After a WebSphere Liberty application is exposed as a Provisioned Service, a service binding request can refer to the application as a backing service.

The instructions that follow show how to bind WebSphere Liberty applications as services or producers to other workloads (such as pods or deployments). Two WebSphere Liberty applications that are deployed through the WebSphere Liberty operator cannot be bound. For more information, see Known issues and limitations.

• Set up the Service Binding operator to access WebSphere Liberty applications.
  By default, the Service Binding operator does not have permission to interact with WebSphere Liberty applications that are deployed through the WebSphere Liberty operator. You must create two RoleBindings to give the Service Binding operator view and edit access for WebSphere Liberty applications.

  1. In the Red Hat OpenShift dashboard, navigate to User Management > RoleBindings.
  2. Select Create binding.
  3. Set the Binding type to Cluster-wide role binding (ClusterRoleBinding).
  4. Enter a name for the binding. Choose a name that is related to service bindings and view access for WebSphere applications.
  5. For the role name, enter webspherelibertyapplications.liberty.websphere.ibm.com-v1-view.
  6. Set the Subject to ServiceAccount.
  7. A Subject namespace menu appears. Select openshift-operators.
  8. In the Subject name field, enter service-binding-operator.
  9. Click Create.
  Now that you have set up the first role binding, navigate to the RoleBindings list and click Create binding again. Set up edit access by using the following instructions.

  1. Set Binding type to Cluster-wide role binding (ClusterRoleBinding).
  2. Enter a name for the binding. Choose a name that is related to service bindings and edit access for WebSphere applications.
  3. In the Role name field, enter webspherelibertyapplications.liberty.websphere.ibm.com-v1-edit.
  4. Set Subject to ServiceAccount.
  5. In the Subject namespace list, select openshift-operators.
  6. In the Subject name field, type service-binding-operator.
  7. Click Create.

  Service bindings from WebSphere Liberty applications (or "services") to pods or deployments (or "workloads") now succeed. After a binding is made, the bound workload restarts or scales to mount the binding secret to /bindings in all containers.

• Set up a service binding by using the Red Hat method.
  For more information, see the Red Hat documentation or the Red Hat tutorial.

  1. On the Red Hat OpenShift web dashboard, click Administrator in the sidebar and select Developer.
  2. In the Topology view for the current namespace, hover over the border of the WebSphere application to be bound as a service, and drag an arrow to the Pod or Deployment workload. A tooltip appears entitled Create Service Binding.
  3. The Create Service Binding window opens. Change the name to value that is fewer than 63 characters. The Service Binding operator might fail to mount the secret as a volume if the name exceeds 63 characters.
  4. Click Create.
  5. A sidebar opens. To see the status of the binding, click the name of the secret and then scroll until the status appears.
  6. Check the pod/deployment workload and verify that a volume is mounted. You can also open a terminal session into a container and run ls /bindings.
• Set up a service binding using the Spec API Tech Preview / Community method.
  This method is newer than the Red Hat method but achieves the same results. You must add a label to your WebSphere Liberty application, such as app=frontend, if it does not have any unique labels. Set the binding to use a label selector so that the Service Binding operator looks for a WebSphere Liberty application with a specific label.

  1. Install the Service Binding operator by using the Red Hat OpenShift Operator Catalog.
  2. Select Operators > Installed Operators and set the namespace to the same one used by both your WebSphere application and pod/deployment workload.
  3. Open the Service Binding (Spec API Tech Preview) page.
  4. Click Create ServiceBinding.
  5. Choose a short name for the binding. Names that exceed 63 characters might cause the binding secret volume mount to fail.
  6. Expand the Service section.
  7. In the Api Version field, enter liberty.websphere.ibm.com/v1.
  8. In the Kind field, enter WebSphereLibertyApplication.
  9. In the Name field, enter the name of your application. You can get this name from the list of applications on the WebSphere Liberty operator page.
  10. Expand the Workload section.
  11. Set the Api Version field to the value of apiVersion in your target workload YAML. For example, if the workload is a deployment, the value is apps/v1.
  12. Set the Kind field to the value of kind in your target workload YAML. For example, if the workload is a deployment, the value is Deployment.
  13. Expand the Selector subsection, and then expand the Match Expressions subsection.
  14. Click Add Match Expression.
  15. In the Key field, enter the label key that you set earlier. For example, for the label app=frontend, the key is app).
  16. In the Operator field, enter Exists.
  17. Expand the Values subsection and click Add Value.
  18. In the Value field, enter the label value that you set earlier. For example, if using the label app=frontend, the value is frontend.
  19. Click Create.
  20. Check the Pod/Deployment workload and verify that a volume is mounted, either by scrolling down or by opening a terminal session into a container and running ls /bindings.

Limit a pod to run on specified nodes (.spec.affinity)

Use .spec.affinity to constrain a Pod to run only on specified nodes.

To set required labels for pod scheduling on specific nodes, use the .spec.affinity.nodeAffinityLabels field.

apiVersion: liberty.websphere.ibm.com/v1
Kind: WebSphereLibertyApplication
metadata:
  name: my-app
  namespace: test
spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  affinity:
    nodeAffinityLabels:
      customNodeLabel: label1, label2
      customNodeLabel2: label3

The following example requires a large node type and preferences for two zones, which are named zoneA and zoneB.

apiVersion: liberty.websphere.ibm.com/v1
Kind: WebSphereLibertyApplication
metadata:
  name: my-app
  namespace: test
spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  affinity:
    nodeAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
        nodeSelectorTerms:
        - matchExpressions:
          - key:  node.kubernetes.io/instance-type
            operator: In
            values:
            - large
      preferredDuringSchedulingIgnoredDuringExecution:
      - weight: 60
        preference:
          matchExpressions:
          - key: failure-domain.beta.kubernetes.io/zone
            operator: In
            values:
            - zoneA
      - weight: 20
        preference:
          matchExpressions:
          - key: failure-domain.beta.kubernetes.io/zone
            operator: In
            values:
            - zoneB

Use pod affinity and anti-affinity to constrain which nodes your pod is eligible to be scheduled based on labels on pods that are already running on the node rather than based on labels on node.

The following example shows that pod affinity is required and that the pods for Service-A and Service-B must be in the same zone. Through pod anti-affinity, it is preferable not to schedule Service_B and Service_C on the same host.

apiVersion: liberty.websphere.ibm.com/v1
Kind: WebSphereLibertyApplication
metadata:
  name: Service-B
  namespace: test
spec:
  applicationImage: quay.io/my-repo/my-app:1.0
  affinity:
    podAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
      - labelSelector:
          matchExpressions:
          - key: service
            operator: In
            values:
            - Service-A
        topologyKey: failure-domain.beta.kubernetes.io/zone
    podAntiAffinity:
      preferredDuringSchedulingIgnoredDuringExecution:
      - weight: 100
        podAffinityTerm:
          labelSelector:
            matchExpressions:
            - key: service
              operator: In
              values:
              - Service-C
          topologyKey: kubernetes.io/hostname

Constrain how pods are spread between nodes and zones (.spec.topologySpreadConstraints)

Use the .spec.topologySpreadConstraints YAML object to specify constraints on how pods of the application instance (and if enabled, the Semeru Cloud Compiler instance) are spread between nodes and zones of the cluster.

Using the .spec.topologySpreadConstraints.constraints field, you can specify a list of Pod TopologySpreadConstraints to be added, such as in the example below:

apiVersion: liberty.websphere.ibm.com/v1
Kind: WebSphereLibertyApplication
metadata:
  name: my-app
  namespace: test
spec:
  topologySpreadConstraints:
    constraints:
      - maxSkew: 1
        topologyKey: kubernetes.io/hostname
        whenUnsatisfiable: ScheduleAnyway
        labelSelector:
          matchLabels:
            app.kubernetes.io/instance: my-app

By default, the operator will add the following Pod topology spread constraints on the application instance's pods (and if applicable, the Semeru Cloud Compiler instance's pods). The default behaviour is to constrain the spread of pods which are owned by the same application instance (or Semeru Cloud Compiler generation instance), denoted by <instance name> with a maxSkew of 1.

- maxSkew: 1
  topologyKey: kubernetes.io/hostname
  whenUnsatisfiable: ScheduleAnyway
  labelSelector:
    matchLabels:
      app.kubernetes.io/instance: <instance name>
- maxSkew: 1
  topologyKey: topology.kubernetes.io/zone
  whenUnsatisfiable: ScheduleAnyway
  labelSelector:
    matchLabels:
      app.kubernetes.io/instance: <instance name>

To remove the operator's default topology spread constraints from above, set the .spec.topologySpreadConstraints.disableOperatorDefaults flag to true.

apiVersion: liberty.websphere.ibm.com/v1
Kind: WebSphereLibertyApplication
metadata:
  name: my-app
  namespace: test
spec:
  topologySpreadConstraints:
    disableOperatorDefaults: true

Alternatively, override each constraint manually by creating a new TopologySpreadConstraint under .spec.topologySpreadConstraints.constraints for each topologyKey you want to modify.

Configure DNS (.spec.dns.policy and .spec.dns.config)

DNS can be configured in WebSphereLibertyApplication CR by using the .spec.dns.policy field or the .spec.dns.config field. The .spec.dns.policy field is the DNS policy for the application pod and defaults to the ClusterFirst policy. The .spec.dns.config field is the DNS config for the application pod.

DNS Config allows users more control over the DNS settings for an application Pod.

The .spec.dns.config field is optional and it can work with any .spec.dns.policy settings. However, when a .spec.dns.policy is set to None, the .spec.dns.config field must be specified.

spec:
  dns:
    policy: "None"
    config:
      nameservers:
        - 192.0.2.1 # this is an example
      searches:
        - ns1.svc.cluster-domain.example
        - my.dns.search.suffix
      options:
        - name: ndots
          value: "2"
        - name: edns0

For more information on DNS, see the Kubernetes DNS documentation.

Configure tolerations (.spec.tolerations)

Node affinity is a property that attracts pods to a set of nodes either as a preference or a hard requirement. However, taints allow a node to repel a set of pods.

Tolerations are applied to pods and allow a scheduler to schedule pods with matching taints. The scheduler also evaluates other parameters as part of its function.

Taints and tolerations work together to help ensure that application pods are not scheduled onto inappropriate nodes. If one or more taints are applied to a node, the node cannot accept any pods that do not tolerate the taints.

Tolerations can be configured in WebSphereLibertyApplication CR by using the .spec.tolerations field.

spec:
  tolerations:
  - key: "key1"
    operator: "Equal"
    value: "value1"

For more information on taints and toleration, see the Kubernetes taints and toleration documentation.