This project contains multiple Zeebe spring boot workers that implement the service tasks required in the hello-world.bpm process.

service1 - deploys the bpmn diagram from service1/src/main/resources/hello-world.bpmn, implements the Hello service task

service2 - implements the World service task

• Docker compose 1.29
• Download zeeba-hazelcast-exporter fat jar to the root of this project:

wget -O zeebe-hazelcast-exporter-1.1.0-jar-with-dependencies.jar https://github.com/camunda-community-hub/zeebe-hazelcast-exporter/releases/download/1.1.0/zeebe-hazelcast-exporter-1.1.0-jar-with-dependencies.jar

The provider docker-compose.yml will spin up Zeebe, Elasticsearch and the Simple Monitor.

Zeebe is configured with the Hazelcast exporter in memory exporter, settings are in the application.yaml

docker-compose --profile in-memory up

Simple Monitor is running at http://localhost:8085

Once the Camunda stack is up and running. Launch the 2 Spring boot applications com.github.thar.camunda.Service1Application and com.github.thar.camunda.Service2Application

Trigger the hello world process either from Simple Monitor or the Camunda Modeler

The process requires an input variable called source , the json value can be anything e.g: source = {"name": "John"}

The experimental folder contains docker-compose files for exporting data from Zeebe to Opensearch instead of Elasticsearch.

The experiment/cluster folder contains the docker equivalent of the Zeebe cluster Fargate stack described below.

Note: CDK v2 is required.

The cdk module contains a CDK application with a number of stacks for deploying Zeebe 8 on AWS Fargate.

$ cdk ls

zeebe-cluster
zeebe-cluster-with-monitor
zeebe-single


$ cdk diff zeebe-cluster

$ cdk deploy zeebe-cluster

zeebe-single will create a single Zeebe service (1 task, 1 container) instance on Fargate. The default VPC is used, the container is deployed into a public subnet. This uses the camunda/zeebe:8.0.0 from Docker hub. There is no additional configuration applied apart from the environment variables

zeebe-cluster, will deploy 1 Zeebe Gateway with 3 Zeebe broker nodes as Fargate services. The nodes use epheneral task storage instead of an EFS file system volume. Service discovery is enabled with a private dns namespace, so nodes can be addressed using <broker-name>.zeebe-cluster:<port>. This is important in order to get the Zeebe cluster to form, the dns names are used as the values passed to ZEEBE_BROKER_CLUSTER_INITIALCONTACTPOINTS. enableDnsSupport and enableDnsHostnames need to be enabled on the VPC - they usually are for the default VPC, but will need to be set if using a different VPC.

zeebe-cluster-with-monitor, same as zeebe-cluster, but also includes the Simple Nonitor application deployed behind an Application Load Balancer

zeebe-hazelcast-stack, will deploy Zeebe with the same network infrastructure as zeebe-stack but will also enable Hazelcast exporter. The Hazelcast exporter is added to custom Docker image and it is saved as a tar in the root project. The docker image will be uploaded to an ECR registry when cdk deploy is run The build-docker.sh script will generate the image and tar file.

$ ./build-docker.sh

$ cdk deploy zeebe-hazelcast-stack

Destroy all resources:

$ cdk destroy

AWS BILL WARNING - The ECR registry created with the zeebe-hazlecast-stack does not get deleted with cdk destroy. This needs to be done manually