Install the helm chart for Flink on Kubernetes. See Flink on Kubernetes for more details.
Make sure local Docker has been assigned at least 8 cores and 8G memory.
-
Point docker-cli to Docker running in minikube
eval $(minikube -p minikube docker-env)
-
Start local Docker registry in minikube Docker (the
docker
command is talking to minikube Docker)docker run -d -p 5001:5000 --restart=always --name registry registry:2
We tunnel port
5001
to5000
, you can choose any available port.
We need to mount a local path on the host for all containers in minikube to work with.
mkdir /tmp/minikubedata
minikube mount /tmp/minikubedata:/data &
We need a custom docker image that extends the flink base image and adds the hudi flink example jar to it. You can build this docker image by running the following command:
./gradlew :example:dockerPushImage
This should have pushed the docker image to the docker registry. You can verify this by running
docker images
This should show you the docker image hudi-flink:latest
in the list of images.
REPOSITORY TAG IMAGE ID CREATED SIZE
localhost:5001/hudi/hudi-flink latest 87b936181d74 32 minutes ago 1.05GB
Create Minio server in Kubernetes.
./gradlew example:createMinioSvc
This will create a pod in Kubernetes that runs the Minio server. You can verify this by running
./gradlew example:openMinioUI
and then opening the Minio UI in your browser by hitting on http://localhost:9090
.
Use the default credentials: minioadmin
and minioadmin
to log into the console.
From the UI, manually create a bucket named test
.
./gradlew example:deleteMinioSvc
./gradlew example:replaceMinioSvc
We can now submit the Flink job to the Flink cluster running in Kubernetes.
./gradlew example:createFlinkApp
This will create a pod in Kubernetes that runs the Flink job. You can verify this by running
./gradlew example:openFlinkUI
and then opening the Flink UI in your browser by hitting on http://localhost:8081
.
./gradlew example:deleteFlinkApp
./gradlew example:replaceFlinkApp