NOTE: This project is deprecated. Please refer to the community official repo https://github.com/apache/flink-kubernetes-operator
Flink native Kubernetes Operator is a control plane for running Apache Flink native application on Kubernetes. It is java
implemented operator, which use fabric8 Kubernetes Client to contact with K8s apiserver.
Also instead of running shell commands flink run/run-application
to start a Flink application, we choose to use Flink
client interfaces to do this(e.g. ApplicationDeployer
). Same for the savepoint and job status. Benefit from this, the operator will
have better performance to launch multiple applications.
mvn clean install
- Make Sure that FlinkApplication Custom Resource Definition is already applied onto the cluster. The CRD could be find here. If not, issue the following commands to apply:
kubectl apply -f deploy/crd.yaml
- Build Docker Image
docker build . -t flink-native-k8s-operator:1.0.x
docker push
- Start flink-native-k8s-operator deployment
kubectl apply -f deploy/flink-native-k8s-operator.yaml
- Apply the RBAC for flink
A new ServiceAccount
"flink" will be created with enough permission to create/delete pods and ConfigMaps.
kubectl apply -f deploy/flink-rbac.yaml
- Create a new Flink application
The flink-native-k8s-operator will watch the CRD resources and submit a new Flink application once the CR is applied.
kubectl apply -f deploy/cr.yaml
- Get/List Flink applications Get all the Flink applications running in the K8s cluster
kubectl get flinkapp
Describe a specific Flink application to show the status(including job status, savepoint, ect.)
kubectl describe flinkapp {app_name}
- Delete a Flink application
kubectl delete -f deploy/cr.yaml
OR
kubectl delete flinkapp {app_name}
By default, we expose the JobManager rest port with ClusterIP
, which means it could only be accessed in the cluster. In
order to access the webUI outside of the K8s cluster, the operator will try to create an ingress entry for each application.
Then you could use http://{app_name}.flink.k8s.io for the JobManager webUI.
You should add {app_name}.flink.k8s.io {ingress_ip}
to your local /etc/hosts file.
kubectl get ingress flink-native-k8s-operator
could be used to get the ingress ip address.