The Flink SBT template image serves as a base image to build your own SBT application to run on a Flink cluster. See big-data-europe/docker-flink README for a description how to setup a Flink cluster.
You can build and launch your application on a Flink cluster by extending this image with your sources. The template uses SBT as build tool, so make sure you have a build.sbt file for your application specifying all the dependencies.
The SBT assembly command must create an assembly JAR (or 'uber' JAR) containing your code and its dependencies. Flink and Hadoop dependencies should be listes as provided.
- Create a Dockerfile in the root folder of your project (which also contains a
build.sbt) - Extend the Flink SBT template Docker image
- Configure the following environment variables (unless the default value satisfies):
FLINK_APPLICATION_JAR_NAME(default: application-1.0)FLINK_APPLICATION_MAIN_CLASS(default: my.main.Job)FLINK_APPLICATION_ARGS(default: "")
- Build and run the image
docker build --rm=true -t bde/flink-app .
docker run --name my-flink-app --link flink-master:flink-master -d bde/flink-app
The sources in the project folder will be automatically added to /usr/src/app if you directly extend the Flink SBT template image. Otherwise you will have to add and package the sources by yourself in your Dockerfile with the commands:
COPY . /usr/src/app
RUN cd /usr/src/app \
&& sbt clean assembly
If you overwrite the template's CMD in your Dockerfile, make sure to execute the /template.sh script at the end.
FROM bde2020/flink-sbt-template:1.14.5-hadoop3.2
MAINTAINER Gezim Sejdiu <[email protected]>
ENV FLINK_APPLICATION_JAR_NAME my-app-1.0-SNAPSHOT-with-dependencies
ENV FLINK_APPLICATION_MAIN_CLASS my.main.Job
ENV FLINK_APPLICATION_ARGS "arg1 arg2"
See flink-starter.