Table of Contents generated with DocToc
Operator for deploying KEDA (Kubernetes Event-driven Autoscaling) controller on OpenShift or any Kubernetes cluster with Operator Lifecycle Manager framework installed.
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
- On Operator Hub Marketplace locate and install
KEDA
operator. Choose a namespace where the operator will be installed. Thekeda
namespace is recommended. - Create
KedaController
resource in namespace where the operator was installed (e.g.keda
)
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 installation of KEDA is triggered by the creation of
a KedaController
custom resource.
Only custom resource named keda
in the namespace where the operator was
installed (typically, keda
) will trigger the installation, reconfiguration,
or removal of the KEDA Controller resources.
The operator will behave in this manner whether it is installed with the
AllNamespaces
or OwnNamespace
install mode. While the operator more
closely matches the OwnNamespace
semantics, AllNamespaces
is a
supported installation mode to allow it to be installed to namespaces with
existing OperatorGroups
which require that installation mode.
There should be only one KEDA Controller in the cluster.
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 where the operator is installed (usually '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
## CA Certificate ConfigMap Names
# ConfigMaps containing PEM-encoded trusted certificate authorities (CAs).
# The files from the ConfigMaps will be loaded by the KEDA operator during
# start-up and will be used by scalers to authenticate TLS-enabled metrics
# data sources.
# default value: []
# caConfigMaps: []
## 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/
# resources:
# 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"
# --- Audit Config Example 1 ---
## Log request metadata but not request or response body to stdout
# auditConfig:
# policy:
# rules:
# - level: Metadata
# --- Audit Config Example 2 ---
## Log request metadata to PersistentVolumeClaim with max output file size of 50MB
# auditConfig:
# logOutputVolumeClaim: "persistentVolumeClaimName"
# policy:
# rules:
# - level: Metadata
# lifetime:
# maxSize: "50"
# --- Audit Config Example 3 ---
## Omits all requests in RequestReceived stage, first rule logs pod changes
## at RequestResponse level & second rule forbids logging requests to a
## configmap called "controller-leader".
# auditConfig:
# policy:
# omitStages:
# - "RequestReceived"
# rules:
# - level: RequestResponse
# resources:
# - group: ""
# resources: ["pods"]
# - level: None
# resources:
# - group: ""
# resources: ["configmaps"]
# resourceNames: ["controller-leader"]
## 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/
# resources:
# requests:
# cpu: 100m
# memory: 100Mi
# limits:
# cpu: 1000m
# memory: 1000Mi
## KEDA Admission Webhooks related config
admissionWebhooks:
## Logging level for KEDA Admission Webhooks
# allowed values: 'debug', 'info', 'error', or an integer value greater than 0, specified as string
# default value: info
logLevel: info
## Logging format for KEDA Admission Webhooks
# allowed values are json and console
# default value: console
logEncoder: console
## Logging time encoding for KEDA Admission Webhooks
# 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 Admission Webhooks Deployment
# https://kubernetes.io/docs/concepts/overview/working-with-objects/annotations/
# deploymentAnnotations:
# annotationKey: annotationValue
## Labels to be added to the KEDA Admission Webhooks Deployment
# https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
# deploymentLabels:
# labelKey: labelValue
## Annotations to be added to the KEDA Admission Webhooks Pod
# https://kubernetes.io/docs/concepts/overview/working-with-objects/annotations/
# podAnnotations:
# annotationKey: annotationValue
## Labels to be added to the KEDA Admission Webhooks Pod
# https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
# podLabels:
# labelKey: labelValue
## Node selector for pod scheduling for KEDA Admission Webhooks
# https://kubernetes.io/docs/concepts/scheduling-eviction/assign-pod-node/
# nodeSelector:
# beta.kubernetes.io/os: linux
## Tolerations for pod scheduling for KEDA Admission Webhooks
# 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 Admission Webhooks
# 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 Admission Webhooks
# https://kubernetes.io/docs/concepts/configuration/pod-priority-preemption/
# priorityClassName: high-priority
## Manage resource requests & limits for KEDA Admission Webhooks
# https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/
# resources:
# 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
Locate installed KEDA
Operator in keda
namespace and then remove created KedaController
resource or simply delete the KedaController
resource:
kubectl delete -n keda -f config/samples/keda_v1alpha1_kedacontroller.yaml
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
This operator contains monitoring configuration to enable Prometheus metrics collection. ServiceMonitor and PodMonitor instances are created if the CRDs from the Monitoring API are available in the cluster.
This project uses the following tools for development.
To install golangci-lint
locally follow the official documentation.
To install pre-commit
locally follow the official documentation.
pre-commit
uses the .pre-commit-config.yaml configuration file located in the root of the
project.
To set up the git
hook script execute the following command so that the pre-commit steps runs automatically on
each commit.
pre-commit install
This operator was created using the operator-sdk. And uses Operator Lifecycle Manager to describe deployment metadata.
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
To build the operator:
make build