openshift/custom-metrics-autoscaler-operator

Operator to deploy Custom Metrics Autoscaler for OpenShift

Table of Contents generated with DocToc

• KEDA OLM Operator
  • Installation
    • Operator Hub Installation
    • Manual installation
  • The KedaController Custom Resource
    • KedaController Spec
  • Uninstallation
    • How to uninstall KEDA Controller
    • How to uninstall KEDA OLM Operator
  • Development
    • Operator Framework
    • Running locally
    • Building the Operator Image

KEDA OLM Operator

Operator for deploying KEDA controller on OpenShift or any Kubernetes cluster with Operator Lifecycle Manager framework installed.

Installation

Please note that you can not run both KEDA v1 and v2 on the same Kubernetes cluster. You need to uninstall KEDA v1 first, in order to install and use KEDA v2. Don't forget to uninstall KEDA v1 CRDs as well, to ensure that, please run:

kubectl delete crd scaledobjects.keda.k8s.io
kubectl delete crd triggerauthentications.keda.k8s.io

Operator Hub Installation

1. On Operator Hub Marketplace locate and install KEDA operator
2. Create KedaController resource in keda namespace

Manual installation

The following will install KEDA and configure it appropriately for your cluster, please run these commands:

make deploy                                                                  # deploy KEDA OLM Operator
kubectl apply -n keda -f config/samples/keda_v1alpha1_kedacontroller.yaml    # install KEDA

To be clear, the operator will be deployed in the keda namespace, and then it will install KEDA into this namespace.

The KedaController Custom Resource

The installation of KEDA is triggered by the creation of a KedaController custom resource. Only custom resource named keda in namespace keda will trigger the installation, reconfiguration, or removal of the KEDA Controller resources.

There could be only one KEDA Controller in the cluster.

KedaController Spec

apiVersion: keda.sh/v1alpha1
kind: KedaController
metadata:
  name: keda
  namespace: keda
spec:
  ###
  # THERE SHOULD BE ONLY ONE INSTANCE OF THIS RESOURCE PER CLUSTER
  # with Name set to 'keda' created in namespace 'keda'
  ###

  ## Namespace that should be watched by KEDA,
  # omit or set empty to watch all namespaces (default setting)
  watchNamespace: ""

  ## KEDA Operator related config
  operator:
    ## Logging level for KEDA Operator
    # allowed values: 'debug', 'info', 'error', or an integer value greater than 0, specified as string
    # default value: info
    logLevel: info

    ## Logging format for KEDA Operator
    # allowed values are json and console
    # default value: console
    logEncoder: console

    ## Logging time encoding for KEDA Controller
    # allowed values are 'epoch', 'millis', 'nano', 'iso8601', 'rfc3339' or 'rfc3339nano'
    # default value: rfc3339
    # logTimeEncoding: rfc3339

    ## Arbitrary arguments
    # Define any argument with possibility to override already existing ones.
    # Array of strings (format is either with prefix '--key=value' or just 'value')
    # args: []

    ## Annotations to be added to the KEDA Operator Deployment
    # https://kubernetes.io/docs/concepts/overview/working-with-objects/annotations/
    # deploymentAnnotations:
    #  annotationKey: annotationValue

    ## Labels to be added to the KEDA Operator Deployment
    # https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
    # deploymentLabels:
    #  labelKey: labelValue

    ## Annotations to be added to the KEDA Operator Pod
    # https://kubernetes.io/docs/concepts/overview/working-with-objects/annotations/
    # podAnnotations:
    #  annotationKey: annotationValue

    ## Labels to be added to the KEDA Operator Pod
    # https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
    # podLabels:
    #  labelKey: labelValue

    ## Node selector for pod scheduling for KEDA Operator
    # https://kubernetes.io/docs/concepts/scheduling-eviction/assign-pod-node/
    # nodeSelector:
    #  beta.kubernetes.io/os: linux

    ## Tolerations for pod scheduling for KEDA Operator
    # https://kubernetes.io/docs/concepts/scheduling-eviction/taint-and-toleration/
    # tolerations:
    # - key: "key1"
    #   operator: "Equal"
    #   value: "value1"
    #   effect: "NoSchedule"
    # - key: "key1"
    #   operator: "Equal"
    #   value: "value1"
    #   effect: "NoExecute"

    ## Affinity for pod scheduling for KEDA Operator
    # https://kubernetes.io/docs/tasks/configure-pod-container/assign-pods-nodes-using-node-affinity/
    # affinity:
    #  podAntiAffinity:
    #    requiredDuringSchedulingIgnoredDuringExecution:
    #     - labelSelector:
    #         matchExpressions:
    #         - key: app
    #           operator: In
    #           values:
    #           - keda-operator
    #           - keda-operator-metrics-apiserver
    #       topologyKey: "kubernetes.io/hostname"

    ## Pod priority for KEDA Operator
    # https://kubernetes.io/docs/concepts/configuration/pod-priority-preemption/
    # priorityClassName: high-priority

    ## Manage resource requests & limits for KEDA Operator
    # https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/
    # resourcesKedaOperator:
    #   requests:
    #     cpu: 100m
    #     memory: 100Mi
    #   limits:
    #     cpu: 1000m
    #      memory: 1000Mi

  ## KEDA Metrics Server related config
  metricsServer:
    ## Logging level for Metrics Server
    # allowed values: "0" for info, "4" for debug, or an integer value greater than 0, specified as string
    # default value: "0"
    logLevel: "0"

    ## Arbitrary arguments
    # Define any argument with possibility to override already existing ones.
    # Array of strings (format is either with prefix '--key=value' or just 'value')
    # args: []

    ## Audit Config
    # https://kubernetes.io/docs/tasks/debug/debug-cluster/audit/#audit-policy
    # Define basic arguments for auditing log files. If needed, more complex flags
    # can be set via 'Args' field manually.
    # Non-empty 'policy' field is mandatory to enable logging.
    # If 'logOutputVolumeClaim' is empty the audit log is printed to stdout,
    # otherwise it points to the user defined PersistentVolumeClaim resource name.
    # auditConfig:
    #   logFormat: "json"
    #   logOutputVolumeClaim: "persistentVolumeClaimName"
    #   policy:
    #     rules:
    #     - level: Metadata
    #     omitStages: "RequestReceived"
    #     omitManagedFields: false
    #   lifetime:
    #     maxAge: "2"
    #     maxBackup: "1"
    #     maxSize: "50"

    ## Annotations to be added to the KEDA Metrics Server Deployment
    # https://kubernetes.io/docs/concepts/overview/working-with-objects/annotations/
    # deploymentAnnotations:
    #  annotationKey: annotationValue

    ## Labels to be added to the KEDA Metrics Server Deployment
    # https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
    # deploymentLabels:
    #  labelKey: labelValue

    ## Annotations to be added to the KEDA Metrics Server Pod
    # https://kubernetes.io/docs/concepts/overview/working-with-objects/annotations/
    # podAnnotations:
    #  annotationKey: annotationValue

    ## Labels to be added to the KEDA Metrics Server Pod
    # https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
    # podLabels:
    #  labelKey: labelValue

    ## Node selector for pod scheduling for Metrics Server
    # https://kubernetes.io/docs/concepts/scheduling-eviction/assign-pod-node/
    # nodeSelector:
    #  beta.kubernetes.io/os: linux

    ## Tolerations for pod scheduling for KEDA Metrics Server
    # https://kubernetes.io/docs/concepts/scheduling-eviction/taint-and-toleration/
    # tolerations:
    # - key: "key1"
    #   operator: "Equal"
    #   value: "value1"
    #   effect: "NoSchedule"
    # - key: "key1"
    #   operator: "Equal"
    #   value: "value1"
    #   effect: "NoExecute"

    ## Affinity for pod scheduling for KEDA Metrics Server
    # https://kubernetes.io/docs/tasks/configure-pod-container/assign-pods-nodes-using-node-affinity/
    # affinity:
    #  podAntiAffinity:
    #    requiredDuringSchedulingIgnoredDuringExecution:
    #     - labelSelector:
    #         matchExpressions:
    #         - key: app
    #           operator: In
    #           values:
    #           - keda-operator
    #           - keda-operator-metrics-apiserver
    #       topologyKey: "kubernetes.io/hostname"

    ## Pod priority for KEDA Metrics Server
    # https://kubernetes.io/docs/concepts/configuration/pod-priority-preemption/
    # priorityClassName: high-priority

    ## Manage resource requests & limits for KEDA Metrics Server
    # https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/
    # resourcesKedaOperator:
    #   requests:
    #     cpu: 100m
    #     memory: 100Mi
    #   limits:
    #     cpu: 1000m
    #      memory: 1000Mi

  ## KEDA ServiceAccount related config
  serviceAccount:
    ## Annotations to be added to the Service Account
    # https://kubernetes.io/docs/concepts/overview/working-with-objects/annotations/
    # annotations:
    #  annotationKey: annotationValue

    ## Labels to be added to the ServiceAccount
    # https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
    # labels:
    #  labelKey: labelValue

Uninstallation

How to uninstall KEDA Controller

Locate installed KEDA Operator in keda namespace and then remove created KedaController resoure or simply delete the KedaController resource:

kubectl delete -n keda -f config/samples/keda_v1alpha1_kedacontroller.yaml

How to uninstall KEDA OLM Operator

To remove KEDA OLM Operator from your cluster, on Operator Hub locate and uninstall KEDA operator.

In case of manual installation, run these commands:

make undeploy

Development

Operator Framework

This operator was created using the operator-sdk. And uses Operator Lifecycle Manager to describe deployment metadata.

Running locally

It can be convenient to run the operator outside of the cluster to test changes. The following command will build the operator and use your current "kube config" to connect to the cluster:

make install    # install KedaController CRD in the cluster
make run        # run operator locally

Building the Operator Image

To build the operator:

make build