Cluster-Curator-Controller

What is cluster-curator-controller

This project contains jobs and controllers for curating work around Cluster Provisioning. It is designed to extend the capabilities already present within Hive ClusterDeployment and Open Cluster Management ManagedCluster kinds. Hosted cluster support was added as well so now the controller also handles HostedCluster and NodePool kinds. The controller will automatically detect between Hive and Hosted clusters and will handle each type accordingly.

Architecture

The controller found here, monitors for ClusterCurator kind. When new instances of the resource are created, this controller creates a Kubernetes Job (curator job) that runs: pre-hook Ansible, activate-and-monitor and posthook Ansible. The Job can be overridden with your own job flow.

For more details on job flow within our architecture see our swimlane chart.

Getting started

• Built in curation jobs:

  Job action	Description	Cloud Provider	Cluster Curator
  applycloudprovider-(aws/gcp/azure/vmware)	Creates AWS/GCP/Azure/VMware related credentials for a cluster deployment	X	X
  applycloudprovider-ansible	Creates the Ansible tower secret for a cluster deployment (included in applycloudprovider-aws)	X	X
  activate-and-monitor	Sets ClusterDeployment.spec.installAttempsLimit: 1, then monitors the deployment of the cluster		X
  monitor-import	Monitors the ManagedCluster import		X
  prehook-ansiblejob posthook-ansiblejob	Creates an AnsibleJob resource and monitors it to completion		X
  monitor	Watches a ClusterDeployment Provisioning Job

  • Here is an example of each job described above. You can add and remove instances of the job containers as needed. You can also inject your own containers ./deploy/jobs/create-cluster.yaml

---

• Steps for development:

  • Compile binaries

    make compile-curator

    This will compile the two binaries manager and curator in ./build/_output

  • Push image to a repo

    • Connect your local docker daemon to your prefered registry

      export VERSION=1.0    # This is used as the docker image tag
      
      export REPO_URL=quay.io/MY_REPOSITORY  # This can also be a docker registry
      
      make push-curator

    • You can now modify the image: references (there are x2) in deploy/controller/deployment.yaml

  • Updating the Type.go and CRD

    • Requires controller-gen be installed, the makefile will attempt to install it if it is not present.

      make generate   # Creates the groups and deep copy scafold
      
      make manifests  # Regenerates the CRDs

---

• Steps for deployment:

  • Connect to the OpenShift hub, the cluster with Open Cluster Management installed

    oc apply -k deploy/controller

  • This deployment defaults to the namespace open-cluster-management. Each time a new ClusterCurator resource is created, you will see operations take place in the controller pod's log, as well as the status.conditions on the ClusterCurator resource.

---

• Hive cluster provisioning example: (AWS)

  • In the Red Hat Advanced Cluster Management console, create a NEW cluster. Before you press the button to create the button, flip the YAML switch. Add the annotation hive.openshift.io/reconcile-pause=true, then press Create

  • The cluster will show in the console as Creating, but it is waiting for curation. If you will be using Ansible, create a secret in the cluster namespace. Here is a an example:

    ---
    apiVersion: v1
    kind: Secret
    metadata:
      name: toweraccess
      namespace: MY_CLUSTER_NAMESPACE
    stringData:
      host: https://my-tower-domain.io
      token: ANSIBLE_TOKEN_FOR_admin  

  • Now create the ClusterCurator resource kind in the cluster namespace. This will begin curation of your cluster provisioning.

    ---
    apiVersion: cluster.open-cluster-management.io/v1beta1
    kind: ClusterCurator
    metadata:
      name: MY_CLUSTER_NAME
      namespace: MY_CLUSTER_NAMESPACE
      labels:
        open-cluster-management: curator
    spec:
      desiredCuration: install
      install:
        prehook:
          - name: Demo Job Template
            extra_vars:
              variable1: something-interesting
              variable2: 2
          - name: Demo Job Template
        posthook:
          - name: Demo Job Template

  • For each ansibleJob that will be created, two special keys will be created in extra_vars:

    1. cluster_deployment, which has general information about the cluster
    2. machine_pool which has details about the worker nodes being created in the cluster

    These are made available to be used by the AnsibleJob during execution

  • The provisiong will start. To find the curator job, run the following command:

    oc -n MY_CLUSTER get ClusterCurator -o yaml

    This will return the YAML for the ClusteCurator resource, in the spec is a field curatorJob, this has the Job that is curating the cluster. Also note the status.conditions, each step performed by the curator job will have a condition where the status: "False". The type of each condition is a step(InitContainer) in the curator job. You can combine the job name and the type value to retreive the logs

    spec:
      curatorJob: curator-job-d9pwh
      ...
    status:
      conditions:
      - lastTransitionTime: "2021-03-30T03:58:59Z"
        message: Executing init container prehook-ansiblejob
        reason: Job_has_finished
        status: "False"
        type: prehook-ansiblejob
    

Hosted cluster provisioning example: (KubeVirt)

• When creating the HostedCluster and NodePool resource add the spec.pausedUntil field with value true to both resources. If using the hcp create cluster CLI you can specify the flag --pausedUntil true.

  apiVersion: hypershift.openshift.io/v1beta1
  kind: HostedCluster
  metadata:
    name: my-cluster
    namespace: clusters
  spec:
    pausedUntil: 'true'
  ...

  apiVersion: hypershift.openshift.io/v1beta1
  kind: NodePool
  metadata:
    name: my-cluster-us-east-2
    namespace: clusters
  spec:
    pausedUntil: 'true'
  ...

• The cluster will show in the console as Creating, but it is waiting for curation. If you will be using Ansible, create a secret in the same namespace as the HostedCluster resource. Here is a an example:

  ---
  apiVersion: v1
  kind: Secret
  metadata:
    name: toweraccess
    namespace: HOSTED_CLUSTER_NAMESPACE
  stringData:
    host: https://my-tower-domain.io
    token: ANSIBLE_TOKEN_FOR_admin  

• Now create the ClusterCurator resource kind in the HostedCluster namespace. The ClusterCurator name is required to be the same as the HostedCluster name. This will begin curation of your cluster provisioning by removing the spec.pausedUntil field in all the HostedCluster and NodePool resources.

  ---
  apiVersion: cluster.open-cluster-management.io/v1beta1
  kind: ClusterCurator
  metadata:
    name: HOSTED_CLUSTER_NAME
    namespace: HOSTED_CLUSTER_NAMESPACE
    labels:
      open-cluster-management: curator
  spec:
    desiredCuration: install
    install:
      prehook:
        - name: Demo Job Template
          extra_vars:
            variable1: something-interesting
            variable2: 2
        - name: Demo Job Template
      posthook:
        - name: Demo Job Template
      towerAuthSecret: toweraccess

Note: The desiredCuration commands are exactly the same as the Hive cluster since the controller will auto-detect the cluster type

• To monitor the status of the provision you can look at either the ClusterCurator status or look at the job logs from the HostedCluster namespace.

---

• Diagnostic steps:

  • Run the following command to see the logs

    # oc logs job/CURATOR_JOB_NAME TYPE_VALUE
    oc logs job/curator-job-d9pwh -c prehook-ansiblejob
    
    oc logs job/curator-job-d9pwh -c activate-and-monitor
    
    oc logs job/curator-job-d9pwh -c posthook-ansiblejob

  • Add a "-f" to the end if you want to tail the output

    If there is a failure, the job will show Failure. Look at the curator-job-container value to see which step in the provisioning failed and review the logs above. If the curator-job-contianer is monitor, there may be an additional provisioning job. Check this log for additional information.

    The generated YAML can be committed to a Git repository. You can then use an ACM Subscription to apply the YAML (provision) on the ACM Hub. Repeat steps 1 & 3 to create new clusters.

---

• Steps for test:

  make unit-tests

• Check the Security guide if you need to report a security issue.