Installing on other Kubernetes platforms

The following sections provide instructions about installing Event Processing on Kubernetes platforms that support the Red Hat Universal Base Images (UBI) containers.

Before you begin

• Ensure you have set up your environment according to the prerequisites.
• Ensure you have planned for your installation, such as preparing for persistent storage, and considering security options.
• Obtain the connection details for your Kubernetes cluster from your administrator.
• If the Kubernetes service Domain Name System (DNS) domain for your cluster is not cluster.local, set kubernetesServiceDnsDomain as required.

Create a namespace

Create a namespace into which the Event Processing instance will be installed. For more information about namespaces, see the Kubernetes documentation.

Ensure you use a namespace that is dedicated to a single instance of Event Processing. This is required because Event Processing uses network security policies to restrict network connections between its internal components. A single namespace per instance also allows for finer control of user accesses.

Important: Do not use any of the initial or system namespaces to install an instance of Event Processing (some examples of these are: default, kube-node-lease, kube-public, and kube-system).

Add the Helm repository

Before you can install the Event Processing operator and the IBM Operator for Apache Flink, add the IBM Helm repository to your local repository list. This will provide access to the Helm chart packages that will install the operators on your cluster.

To add the IBM Helm repository to the local repository list, run the following command:

helm repo add ibm-helm https://raw.githubusercontent.com/IBM/charts/master/repo/ibm-helm

Choose the operator installation mode

Before installing an operator, decide whether you want the operator to:

• Manage instances in any namespace.

  To use this option, select All namespaces on the cluster (default) later. The operator will be deployed into the system namespace openshift-operators, and will be able to manage instances in any namespace.

• Only manage instances in a single namespace.

  To use this option, select A specific namespace on the cluster later. The operator will be deployed into the specified namespace, and will not be able to manage instances in any other namespace.

Important: Choose only one mode when installing the operator. Mixing installation modes is not supported due to possible conflicts. If an operator is installed to manage all namespaces and a single namespace at the same time, it can result in conflicts and attempts to control the same CustomResourceDefinition resources.

Install the CRDs

Before installing the operator, the Custom Resource Definitions (CRDs) for Event Processing and Flink must be installed. To install the CRDs, run the following commands:

For Event Processing:

helm install \
   <release-name> ibm-helm/ibm-ep-operator-crd \
   -n <namespace>

For IBM Operator for Apache Flink:

helm install \
   <release-name> ibm-helm/ibm-eventautomation-flink-operator-crd \
   -n <namespace>

Where:

• <release-name> is the name you provide to identify your Event Processing or Flink CRD Helm installation.
• <namespace> is the name of the namespace where you want to install the Helm release that you use for CRD management.

For example, to install the operator CRD Helm charts in the ep namespace, run the commands as follows:

helm install ep-crds ibm-helm/ibm-ep-operator-crd -n ep
helm install ibm-flink-crds ibm-helm/ibm-eventautomation-flink-operator-crd -n ep

Install operators

Follow the instructions to install the IBM Operator for Apache Flink and the Event Processing operator.

Installing the IBM Operator for Apache Flink operator

Important:

• The IBM Operator for Apache Flink must not be installed on a cluster where you already have an Apache Flink operator installed. This is because the IBM Operator for Apache Flink leverages the Apache Flink CustomResourceDefinition (CRD) resources. These resources cannot be managed by more than one operator at the same time (for more information, see the Operator Framework documentation).
• Before installing the IBM Operator for Apache Flink on a cluster where the Apache Flink operator is already installed, uninstall the Apache Flink operator and the Apache Flink CRDs as described in the Apache Flink documentation.
• Only one version of the IBM Operator for Apache Flink must be installed on a cluster. Installing multiple versions is not supported due to potential conflicts between different versions of the CustomResourceDefinition resources.
• Before you install the IBM Operator for Apache Flink, ensure that you have created truststores and keystores that are required to secure communication with Flink deployments.

Ensure you have considered the IBM Operator for Apache Flink requirements, including resource requirements and, if installing in any namespace, the required cluster-scoped permissions.

To install the operator, run the following command:

helm install \
   <release-name> ibm-helm/ibm-eventautomation-flink-operator \
   -n <namespace> \
   --set kubernetesServiceDnsDomain=<your.k8s.svc.dns.domain>  \
   --set watchAnyNamespace=<true/false>

Where:

• <release-name> is the name you provide to identify your operator.
• <namespace> is the name of the namespace where you want to install the operator.
• kubernetesServiceDnsDomain=<your.k8s.svc.dns.domain> allows certificate to be created by utilizing the correct Kubernetes service DNS domain as a suffix on hosts in the dnsNames section of the certificate (default is cluster.local if not specified).

• watchAnyNamespace=<true/false> determines whether the operator manages FlinkDeployments in any namespace or only a single namespace (default is false if not specified).

  Set to true for the operator to manage instances in any namespace, or do not specify if you want the operator to only manage instances in a single namespace.

For example, to install the operator on a cluster where it will manage all instances of FlinkDeployments, run the command as follows:

helm install \
   flink ibm-helm/ibm-eventautomation-flink-operator\
   -n "my-namespace" \
   --set watchAnyNamespace=true

For example, to install the operator that will manage FlinkDeployments in only the my-flink namespace with no custom configurations such as watchAnyNamespace,kubernetesServiceDnsDomain, or webhook.create, run the command as follows:

helm install flink ibm-helm/ibm-eventautomation-flink-operator -n "my-flink"

Note: If you do not have a Cert Manager installed, disable the webhook component of the Flink operator. To disable the webhooks, specify --set webhook.create=false on the helm install command as follows:

helm install flink ibm-helm/ibm-eventautomation-flink-operator -n "my-flink" --set webhook.create=false

Installing the Event Processing operator

Ensure you have considered the Event Processing operator requirements, including resource requirements and the required cluster-scoped permissions.

Important: You can only install one version of the Event Processing operator on a cluster. Installing multiple versions on a single cluster is not supported due to possible compatibility issues as they share the same Custom Resource Definitions (CRDs), making them unsuitable for coexistence.

To install the operator, run the following command:

helm install \
   <release-name> ibm-helm/ibm-ep-operator \
   -n <namespace> \
   --set webhook.create=<true/false> \
   --set kubernetesServiceDnsDomain=<your.k8s.svc.dns.domain>  \
   --set watchAnyNamespace=<true/false>

Where:

• <release-name> is the name you provide to identify your operator.
• <namespace> is the name of the namespace where you want to install the operator.

• webhook.create=<true/false> determines whether the validating webhook is deployed (default is true if not specified).

  Set to false if you do not have a certificate manager installed and will be creating your own certificates for your Event Processing instances.

• kubernetesServiceDnsDomain=<your.k8s.svc.dns.domain> allows certificate to be created by utilizing the correct Kubernetes service DNS domain as a suffix on hosts in the dnsNames section of the certificate (default is cluster.local if not specified).

• watchAnyNamespace=<true/false> determines whether the operator manages instances of Event Processing in any namespace or only a single namespace (default is false if not specified).

  Set to true for the operator to manage instances in any namespace, or do not specify if you want the operator to only manage instances in a single namespace.

For example, to install the operator on a cluster where it will manage all instances of Event Processing, run the command as follows:

helm install \
   eventprocessing ibm-helm/ibm-ep-operator\
   -n "my-namespace" \
   --set watchAnyNamespace=true \
   --set webhook.create=false

Or to install the operator to manage Event Processing instances in only the my-eventprocesssing namespace with no custom configurations such as watchAnyNamespace,kubernetesServiceDnsDomain, or webhook.create, run the command as follows:

helm install eventprocesssing ibm-helm/ibm-ep-operator -n "my-eventprocesssing"

Checking the operator status

To check the status of the installed operator, run the following command:

For Event Processing:

kubectl get deployments ibm-ep-operator -n <namespace>

For IBM Operator for Apache Flink:

kubectl get deployments flink-kubernetes-operator -n <namespace>

Where:

• <namespace> is the name of the namespace where the operator is installed.

A successful installation will return a result similar to the following with 1/1 in the READY column:

For Event Processing:

NAME                            READY   UP-TO-DATE   AVAILABLE   AGE
ibm-ep-operator                 1/1     1            1           1d2h

For IBM Operator for Apache Flink:

NAME                            READY   UP-TO-DATE   AVAILABLE   AGE
flink-kubernetes-operator       1/1     1            1           7d4h

Install instances

Follow the instructions to install the IBM Operator for Apache Flink and Event Processing instances.

Installing a Flink instance

Instances of Flink can be created after the IBM Operator for Apache Flink is installed. If the operator was installed into a specific namespace, then it can only be used to manage instances of Flink in that namespace.

If the operator was installed for all namespaces, then it can be used to manage instances of Flink in any namespace, including those created after the operator was deployed.

When installing an instance of Flink, ensure you are using a namespace that the Flink operator is managing.

Installing a Flink instance by using the CLI

To install an instance of Flink from the command-line, you must first prepare a FlinkDeployment custom resource configuration in a YAML file.

A number of sample configuration files are available in GitHub, where you can select the GitHub tag for your Flink version, and then go to /cr-examples/flinkdeployment to access the samples. These range from quick start deployments for non-production development to large scale clusters ready to handle a production workload.

Important: All Flink samples except Quick Start use a PersistentVolumeClaim (PVC), which must be deployed manually as described in planning.

To deploy a Flink instance, run the following commands:

1. Prepare a FlinkDeployment custom resource in a YAML file, using the information provided in FlinkDeployment Reference.

   Notes:

   • If the operator webhook has not been disabled, accept the license agreement(spec.flinkConfiguration.license.accept: 'true'), and set the required licensing configuration parameters for your deployment. Do not set the spec.image and spec.flinkVersion fields as they are automatically included by IBM Operator for Apache Flink.

      spec:
        flinkConfiguration:
          license.use: <license-use-value>
          license.license: L-KCVZ-JL5CRM
          license.accept: 'true'
     

     Where <license-use-value> must be either EventAutomationProduction or EventAutomationNonProduction, depending on your deployment.

   • If the operator webhook has been disabled (--set webhook.create=false), in addition to the licensing parameters you configured in the previous step, you must also include the fields spec.image and spec.flinkVersion. The values for the spec.image and spec.flinkVersion fields can be obtained by looking at the values of the IBM_FLINK_IMAGE and IBM_FLINK_VERSION environment variables on the operator pod. You can get these values by running:

     kubectl set env pod/<flink_operator_pod_name> --list -n <flink_operator_namespace> | grep IBM_FLINK
     

     Use the values from the environment variables in your FlinkDeployment:

     spec:
       flinkVersion: "<value-of-IBM_FLINK_VERSION-env-var>"
       image: "<value-of-IBM_FLINK_IMAGE-env-var>"
       flinkConfiguration:
         license.use: <license-use-value>
         license.license: L-KCVZ-JL5CRM
         license.accept: 'true'
     

   • To secure your communication between Flink pods, add the following snippet to the spec.flinkConfiguration section:

     spec:
       flinkConfiguration:
         security.ssl.enabled: 'true'
         security.ssl.truststore: /opt/flink/tls-cert/truststore.jks
         security.ssl.truststore-password: <jks-password>
         security.ssl.keystore: /opt/flink/tls-cert/keystore.jks
         security.ssl.keystore-password: <jks-password>
         security.ssl.key-password: <jks-password>
         kubernetes.secrets: '<jks-secret>:/opt/flink/tls-cert'
     

2. Apply the configured FlinkDeployment custom resource to your target namespace:

   kubectl apply -f <custom-resource-file-path> -n <target-namespace>
   

   For example:

   kubectl apply -f flinkdeployment_demo.yaml -n my-flink
   

3. Wait for the installation to complete.

Installing an Event Processing instance

Instances of Event Processing can be created after the Event Processing operator is installed. If the operator was installed to manage a specific namespace, then it can only be used to manage instances of Event Processing in that namespace. If the operator was installed to manage all namespaces, then it can be used to manage instances of Event Processing in any namespace, including those created after the operator was deployed.

When installing an instance of Event Processing, ensure you are using a namespace that an operator is managing.

Creating an image pull secret

Before installing an Event Processing instance, create an image pull secret called ibm-entitlement-key in the namespace where you want to create an instance of Event Processing. The secret enables container images to be pulled from the registry.

1. Obtain an entitlement key from the IBM Container software library.
2. Click Entitlement keys in the navigation on the left, and click Add new key, or if you have an existing active key available, click Copy to copy the entitlement key to the clipboard.

3. Create the secret in the namespace that will be used to deploy an instance of Event Processing as follows.

   Name the secret ibm-entitlement-key, use cp as the username, your entitlement key as the password, and cp.icr.io as the docker server:

   kubectl create secret docker-registry ibm-entitlement-key --docker-username=cp --docker-password="<your-entitlement-key>" --docker-server="cp.icr.io" -n "<target-namespace>"
   

Note: If you do not create the required secret, pods will fail to start with ImagePullBackOff errors. In this case, ensure the secret is created and allow the pod to restart.

Installing an Event Processing instance by using the CLI

To install an instance of Event Processing from the command line, you must first prepare an eventprocesssing custom resource configuration in a YAML file.

A number of sample configuration files are included in the Helm chart package to base your deployment on. The sample configurations range from a smaller deployment for non-production development or general experimentation to a large deployment ready to handle a production workload.

The sample configurations are also available in GitHub, where you can select the GitHub tag for your Event Processing version, and then go to /cr-examples/eventprocesssing/kubernetes to access the samples for the Kubernetes platforms that support the Red Hat Universal Base Images (UBI) containers.

Note: If experimenting with Event Processing for the first time, the Quick start sample is the smallest and simplest example that can be used to create an experimental deployment. For a production setup, use the Production sample configuration.

More information about these samples is available in the planning section. You can base your deployment on the sample that most closely reflects your requirements and apply customizations as required.

Important: Ensure that the spec.license.accept field in the custom resource YAML is set to true, and that the correct values are selected for the spec.license.license and spec.license.use fields before deploying the Event Processing instance. These values are used for metering purposes and could result in inaccurate charging and auditing if set incorrectly. For more information about the available options, see the licensing reference.

When modifying the sample configuration, ensure that the following fields are updated based on your requirements:

• To secure your communication between Event Processing and Flink pods, identify the secret that contains the same truststore as your Flink deployment and the secret containing the password for this keystore. Then, add the secret to the spec.flink.tls section. For example:

  spec:
    flink:
      tls:
        secretKeyRef:
           key: <key-containing-password-value>
           name: <flink-jks-password-secret>
         secretName: <flink-jks-secret>
  

• The spec.license.accept field in the custom resource YAML is set to true.
• The correct values are selected for the spec.license.use, spec.license.license, and spec.license.metric fields before deploying an Event Processing instance. For information about the right values for your deployment, see the licensing reference.
• authoring.storage.type field is updated as ephemeral or persistent-claim based on your requirements. See configuring to select the correct storage type and other optional specifications such as storage size, root storage path, and secrets.
• authoring.tls.caSecretName or authoring.tls.secretName field is updated based on your requirements. If neither is specified, self-signed certificates are used. See the configuring section for more information.

• spec.authoring.endpoints[] must contain an entry for the ui endpoint:

  • name field is set to ui.

  • host field is updated with a DNS resolvable hostname for accessing the named service, for example:

      spec:
        authoring:
          endpoints:
            - name: ui
              host: qs-ep-ui.mycluster.domain
    

To deploy an Event Processing instance, run the following command:

1. Apply the configured eventprocesssing custom resource to your target namespace:

   kubectl apply -f <custom-resource-file-path> -n <target-namespace>
   

   For example:

   kubectl apply -f production.yaml -n my-namespace
   

2. Wait for the installation to complete.