This is an example of how to run an Apache Flink application in a containerized environment, using either docker compose or kubernetes.

minio, an s3-compatible filesystem, is used for checkpointing.

zookeeper is used for high availability.

Prerequisites

You'll need docker and kubernetes to run this example. It has been tested with Docker Desktop for Mac, Version 2.2.0.3 (42716), with Kubernetes enabled, using the docker-for-desktop context.

This example depends on FLINK-10817, added in Flink 1.8.0.

Build a docker image

To begin, you'll need to build a job-specific docker image by doing these two steps:

make jar
make image

You will need the resulting image for any of the following approaches for running this job.

Run with docker-compose

You can run this job using docker-compose via make run, which is equivalent to

FLINK_JOB=com.ververica.example.StreamingJob \
FLINK_JOB_ARGUMENTS='' \
docker-compose -f docker/docker-compose.yml up -d \
job-cluster minio-service taskmanager miniosetup zoo1

The flink dashboard is then available at http://localhost:8081 and you will find the minio filesystem browser at http://localhost:9000.

make stop

will tear that all down.

Run with kubectl

cd kubernetes
make run

The flink dashboard will be at http://localhost:30081, and the minio browser will be at http://localhost:30090.

To bring it all down, use

make stop

Is this production-ready?

No. At a minimum you should take care of these things:

• imagePullPolicy: Never
• minio and zookeeper should be run as clusters, rather than in single server mode
• use k8s namespaces

Plus the usual getting-your-flink-job-ready-for-production topics, such as:

• UIDs on the stateful operators