Prepare for installation

Introduction

A single operator instance is capable of managing multiple domains in multiple namespaces, depending on how it is configured. A Kubernetes cluster can host multiple operators, but no more than one per namespace.

Before installing an operator, ensure that each of these prerequisite requirements is met:

  1. Check environment
  2. Set up the operator Helm chart access
  3. Inspect the operator Helm chart
  4. Prepare an operator namespace and service account
  5. Prepare operator image
  6. Determine the platform setting
  7. Choose a security strategy
  8. Choose a domain namespace selection strategy
  9. Choose a Helm release name
  10. Be aware of advanced operator configuration options
  11. Special use cases:

Check environment

  1. Review the Operator prerequisites to ensure that your Kubernetes cluster supports the operator.

  2. It is important to keep in mind that some supported environments have additional help or samples that are specific to the operator, or are subject to limitations, special tuning requirements, special licensing requirements, or restrictions. See Supported environments for details.

  3. If your environment doesn’t already have a Kubernetes setup, then see set up Kubernetes.

  4. If it is not already installed, then install Helm. To install an operator, it is required to install Helm in your Kubernetes cluster. The operator uses Helm to create the necessary resources and then deploy the operator in a Kubernetes cluster.

    To check if you already have Helm available, try the helm version command.

    For detailed instructions on installing Helm, see the GitHub Helm Repository.

  5. Optionally, enable Istio.

Set up the operator Helm chart access

Before installing an operator, the operator Helm chart must be made available. The operator Helm chart includes:

  • Pre-configured default values for the configuration of the operator.
  • Helm configuration value settings for fine tuning operator behavior.
  • Commands for deploying (installing) or undeploying the operator.

You can set up access to the operator Helm chart using the chart repository.

  • Use the operator Helm chart repository that is located at https://oracle.github.io/weblogic-kubernetes-operator/charts or in a custom repository that you control.

  • To set up your Helm installation so that it can access the https://oracle.github.io/weblogic-kubernetes-operator/charts repository and name the repository reference weblogic-operator, use the following command, helm repo add <helm-chart-repo-name> <helm-chart-repo-url>:

    $ helm repo add weblogic-operator https://oracle.github.io/weblogic-kubernetes-operator/charts --force-update

  • To verify that a Helm chart repository was added correctly, or to list existing repositories:

    $ helm repo list

    NAME                 URL
weblogic-operator    https://oracle.github.io/weblogic-kubernetes-operator/charts

  • For example, assuming you have named your repository weblogic-operator, simply use weblogic-operator/weblogic-operator in your Helm commands when specifying the chart location.

    • To list the versions of the operator that you can install from the Helm chart repository:

      $ helm search repo weblogic-operator/weblogic-operator --versions

    • For a specified version of the Helm chart and operator, with the helm pull and helm install commands, use the --version <value> option to choose the version that you want, with the latest value being the default.

Inspect the operator Helm chart

You can find out the configuration values that the operator Helm chart supports, as well as the default values, using the helm show command.

$ helm show values weblogic-operator/weblogic-operator

  • Alternatively, you can view most of the configuration values and their defaults in the operator source in the ./kubernetes/charts/weblogic-operator/values.yaml file.

  • The available configuration values are explained by category in the Operator Helm configuration values section of the operator Configuration Reference.

  • Helm commands are explained in more detail here, see Useful Helm operations.

Prepare an operator namespace and service account

Each operator requires a namespace to run in and a Kubernetes service account within the namespace. The service account will be used to host the operator’s security permissions. Only one operator can run in a given namespace.

To simplifies management and monitoring of an operator, Oracle recommends:

  • When possible, create an isolated namespace for each operator which hosts only the operator, does not host domains, and does not host Kubernetes resources that are unrelated to the operator. Sometimes this is not possible, in which case the operator can be configured to manage domains in its own namespace. For more information, see Choose a security strategy and Choose a domain namespace selection strategy.
  • Creating a dedicated service account for each operator instead of relying on the default service account that Kubernetes creates when a new namespace is created.
  • Directly creating a namespace and service account instead of relying on the operator Helm chart installation to create these resources for you.

Here’s an example of each:

$ kubectl create namespace sample-weblogic-operator-ns

$ kubectl create serviceaccount -n sample-weblogic-operator-ns sample-weblogic-operator-sa

In operator installation steps, you will specify the namespace using the --namespace MY-NAMESPACE operator Helm chart configuration setting on the Helm install command line. If not specified, then it defaults to default. If the namespace does not already exist, then Helm will automatically create it (and Kubernetes will create a default service account in the new namespace). If you later uninstall the operator, then Helm will not remove the specified namespace. These are standard Helm behaviors.

Similarly, you will specify the serviceAccount=MY-SERVICE-ACCOUNT operator Helm chart configuration setting on the Helm install command line to specify the service account in the operator’s namespace that the operator will use. If not specified, then it defaults to default. This service account will not be automatically removed when you uninstall an operator.

Prepare operator image

The operator image must be available to all nodes of your Kubernetes cluster.

Locating an operator image

Production-ready operator images for various supported versions of the operator are publicly located in the operator GitHub Container Registry. Operator GitHub container registry images can be directly referenced using an image name similar to ghcr.io/oracle/weblogic-kubernetes-operator:N.N.N where N.N.N refers to the operator version and ghcr.io is the DNS name of the GitHub container registry. You can also optionally build your own operator image, see the Developer Guide.

Default operator image

Each Helm chart version defaults to using an operator image from the matching version. To find the default image name that will be used when installing the operator, see Inspect the operator Helm chart and look for the image value. The value will look something like this, ghcr.io/oracle/weblogic-kubernetes-operator:N.N.N.

Pulling operator image

In most use cases, Kubernetes will automatically download (pull) the operator image, as needed, to the machines on its cluster nodes.

If you want to manually place an operator image in a particular machine’s container image pool, or test access to an image, then call docker pull. For example:

$ docker pull ghcr.io/oracle/weblogic-kubernetes-operator:4.1.8

Note that if the image registry you are using is a private registry that requires an image pull credential, then you will need to call docker login my-registry-dns-name.com before calling docker pull.

Customizing operator image name, pull secret, and private registry

Sometimes, you may want to specify a custom operator image name or an image pull secret. For example, you may want to deploy a different version than the default operator image name or you may want to place an operator image in your own private image registry.

A private image registry requires using a custom image name for the operator where the first part of the name up to the first slash (/) character is the DNS location of the registry and the remaining part refers to the image location within the registry. A private image registry may also require an image pull registry secret to provide security credentials.

  • To reference a custom image name, specify the image= operator Helm chart configuration setting when installing the operator, for example --set "image=my-image-registry.io/my-operator-image:1.0".
  • To create an image pull registry secret, create a Kubernetes Secret of type docker-registry in the namespace where the operator is to be deployed, as described in Specifying imagePullSecrets on a Pod.
  • To reference an image pull registry secret from an operator installation, there are two options:

Determine the platform setting

It is important to set the correct value for the kubernetesPlatform Helm chart configuration setting when installing the operator.

In particular, beginning with operator version 3.3.2, specify the operator kubernetesPlatform Helm chart setting with the value OpenShift when using the OpenShift Kubernetes platform, for example --set "kubernetesPlatform=OpenShift". This accommodates OpenShift security requirements.

For more information, see kubernetesPlatform.

Choose a security strategy

There are three commonly used security strategies for deploying an operator:

  1. Any namespace with cluster role binding enabled
  2. Any namespace with cluster role binding disabled
  3. Local namespace only with cluster role binding disabled

For a detailed description of the operator’s security-related resources, see the operator’s role-based access control (RBAC) requirements, which are documented here.

Any namespace with cluster role binding enabled

If you want to give the operator permission to access any namespace, then, for most use cases, set the enableClusterRoleBinding operator Helm chart configuration setting to true when installing the operator.

For example --set "enableClusterRoleBinding=true". The default for this setting is true.

This is the most popular security strategy.

Any namespace with cluster role binding disabled

If your operator Helm enableClusterRoleBinding configuration value is false, then an operator is still capable of managing multiple namespaces but a running operator will not have privilege to manage a newly added namespace that matches its namespace selection criteria until you upgrade the operator’s Helm release. See Ensuring the operator has permission to manage a namespace.

NOTE: You will need to manually install the operator CRD because enableClusterRoleBinding is not set to true and installation of the CRD requires cluster role binding privileges. See How to manually install the Domain resource custom resource definition (CRD).

Local namespace only with cluster role binding disabled

If you want to limit the operator so that it can access only resources in its local namespace, then:

This may be necessary in environments where the operator cannot have cluster-scoped privileges, such as may happen on OpenShift platforms when running the operator with a Dedicated namespace strategy.

Choose a domain namespace selection strategy

Before installing your operator, choose the value for its domainNamespaceSelectionStrategy Helm chart configuration setting and its related setting (if any). See Choose a domain namespace section strategy.

See Choose a security strategy.

For a description of common namespace management issues, see Common mistakes and solutions. For reference, see WebLogic domain management.

Choose a Helm release name

The operator requires Helm for installation, and Helm requires that each installed operator be assigned a release name. Helm release names can be the same if they are deployed to different namespaces, but must be unique within a particular namespace.

A typical Helm release name is simply weblogic-operator. The operator samples and documentation often use sample-weblogic-operator.

Be aware of advanced operator configuration options

Review the settings in the Configuration Reference for less commonly used advanced or fine tuning Helm chart configuration options that might apply to your particular use case. These include node selectors, node affinity, Elastic Stack integration, the operator REST API, setting operator pod labels, setting operator pod annotations, and Istio.

Special use cases

If applicable, please review the following special use cases.

How to download the Helm chart if Internet access is not available

At a high level, you use helm pull to download a released version of the Helm chart and move it to the machine with no Internet access, so that then you can run helm install to install the operator.

The steps are:

  1. On a machine with Internet access, to download the chart to the current directory, run:
$ helm pull weblogic-operator --repo https://oracle.github.io/weblogic-kubernetes-operator/charts --destination .

For a specified version of the Helm chart, with helm pull and helm install, use the --version <value> option to choose the version that you want, with the latest value being the default. To list the versions of the operator that you can install from the Helm chart repository, run:

$ helm repo add weblogic-operator https://oracle.github.io/weblogic-kubernetes-operator/charts --force-update
$ helm search repo weblogic-operator/weblogic-operator --versions
  1. Move the resulting weblogic-operator-<version>.tgz file to the machine without Internet access on which you want to install the WebLogic Kubernetes Operator.
  2. Run $ tar zxf weblogic-operator-<version>.tgz. This puts the Helm chart files in a local directory ./weblogic-operator.
  3. To create the namespace where the operator will be installed, run $ kubectl create namespace weblogic-operator.

    Creating a dedicated namespace for the operator is the most common approach, but is not always correct or sufficient. For details, see the prerequisite steps starting with Step 3. Inspect the operator Helm chart. Be sure to follow all the previously detailed prerequisite steps, ending at Step 10. Be aware of advanced operator configuration options.

  4. To install the operator, run $ helm install weblogic-operator ./weblogic-operator --namespace weblogic-operator.

How to manually install the Domain resource custom resource definition (CRD)

The Domain resource type is defined by a Kubernetes CustomResourceDefinition (CRD) resource. The domain CRD provides Kubernetes with the schema for operator domain resources and there must be one domain CRD installed in each Kubernetes cluster that hosts the operator. If you install multiple operators in the same Kubernetes cluster, then they all share the same domain CRD.

When does a Domain CRD need to be manually installed?

Typically, the operator automatically installs the CRD for the Domain type when the operator first starts. However, if the operator lacks sufficient permission to install it, then you may choose to manually install the CRD in advance by using one of the provided YAML files. Manually installing the CRD in advance allows you to run the operator without giving it privilege (through Kubernetes roles and bindings) to access or update the CRD or other cluster-scoped resources. This may be necessary in environments where the operator cannot have cluster-scoped security privileges, such as on OpenShift when running the operator with a Dedicated namespace strategy. See Choose a security strategy.

How to manually install a Domain CRD.

To manually install the CRD, first download the operator source.

Assuming you have installed the operator source into the /tmp directory:

$ cd /tmp/weblogic-kubernetes-operator

$ kubectl create -f kubernetes/crd/domain-crd.yaml

How to check if a Domain CRD has been installed.

After the CustomResourceDefinition is installed, either by the operator or using the previous create command, you can verify that the CRD is installed correctly using:

$ kubectl get crd domains.weblogic.oracle

Or, by calling:

$ kubectl explain domain.spec

The kubectl explain call should succeed and list the domain resource’s domain.spec attributes that are defined in the CRD.