Skip to content

This project uses Apache Flink as a stream engine that consumes data from the File system or Kafka brokers and exposes metrics using Prometheus and Grafana, everything deployed on Kubernetes (minikube).

felipegutierrez/explore-flink

Repository files navigation

Java CI with Maven Run Test Suites codecov CodeFactor Codacy Badge CodeQL GitHub issues GitHub closed issues Docker Image Size (latest by date) Lines of code

This project is based on Apache Flink 1.13.1 consuming events from Kafka 2.2.5 (using Strimzi operators) with Docker 19.03.8, Kubernetes v1.19.0, minikube v1.13.1, Java 8, and Scala 2.12. The docker images can be found at Docker Hub.

1. Kubernetes + Docker + Kafka & Zookeeper (3 brokers & 3 zookeepers from Strimzi operators) + Flink(1 JobManager & 3 TaskManagers) + Prometheus + Grafana

This section aims to use Flink consuming data from Kafka and from a filesystem, exporting data to Promethues and displaying it at Grafana dashboard. It is based on the official tutorial Flink with Kubernetes Setup.

minikube start --cpus 4 --memory 8192

First configure Kafka using the Strimzi operators based on this file. Then, apply the other yaml files available on directory k8s/.

minikube ssh 'sudo ip link set docker0 promisc on'
kubectl proxy
kubectl apply -n kafka -f k8s/

Use the minikube IP address minikube ip to access the Flink UI-Web at http://172.17.0.2:30081, the Prometheus WebUI at http://172.17.0.2:30091/targets and http://172.17.0.2:30091/graph, and the Grafana WebUI at http://172.17.0.2:30011.

Overview of this project in action

Kafka(Strimzi) - Flink web UI - prometheus - grafana - using Kubernetes

Troubleshooting

Testing Kafka client:

$ kubectl exec -it kafka-client -- /bin/bash
root@kafka-client:/# cd /usr/bin/
root@kafka-client:/usr/bin# kafka-topics --list --zookeeper zookeeper:2181
__confluent.support.metrics

List the objects, the resources:

kubectl get all
kubectl api-resources

Logs:

kubectl describe pod <POD_ID>
kubectl logs <POD_ID>
kubectl exec -i -t <POD_ID> -- /bin/bash
kubectl -n kube-system top pods
kubectl get nodes
kubectl top nodes
kubectl get pods -A
kubectl top pods

Clean your Kubernetes cluster and delete everything when you finish to test.

kubectl delete statefulset flink-taskmanager kafka zookeeper
kubectl delete jobs flink-jobmanager tpch-dbgen-job
kubectl delete deployments kafka-broker grafana-deployment prometheus-deployment tpch-dbgen-deployment
kubectl delete services zookeeper kafka flink-jobmanager flink-jobmanager-rest flink-taskmanager prometheus-service-rest prometheus-service grafana-service-rest grafana-service
kubectl delete pvc tpch-dbgen-data-pvc tpch-dbgen-datarate-pvc
kubectl delete configmaps flink-config grafana-config prometheus-config
kubectl delete pods kafka-client

minikube stop

2. Docker + Kafka(1 ZooKeeper, 1 broker) + Flink(1JobManager, 3TaskManagers) + Prometheus + Grafana

This section is here only to help if one would like to not use Kubernetes and deploy Flink only using Docker.

cd operations-playground
docker-compose build --no-cache
docker-compose up -d --remove-orphans
docker-compose ps -a
                    Name                                  Command               State                   Ports                
-----------------------------------------------------------------------------------------------------------------------------
operations-playground_clickevent-generator_1   /docker-entrypoint.sh java ...   Up       6123/tcp, 8081/tcp                  
operations-playground_client_1                 /docker-entrypoint.sh flin ...   Exit 0                                       
operations-playground_jobmanager_1             /docker-entrypoint.sh jobm ...   Up       6123/tcp, 0.0.0.0:8081->8081/tcp    
operations-playground_kafka_1                  start-kafka.sh                   Up       0.0.0.0:9094->9094/tcp              
operations-playground_taskmanager-01_1         /docker-entrypoint.sh task ...   Up       6123/tcp, 8081/tcp                  
operations-playground_taskmanager-02_1         /docker-entrypoint.sh task ...   Up       6123/tcp, 8081/tcp                  
operations-playground_taskmanager-03_1         /docker-entrypoint.sh task ...   Up       6123/tcp, 8081/tcp                  
operations-playground_zookeeper_1              /bin/sh -c /usr/sbin/sshd  ...   Up       2181/tcp, 22/tcp, 2888/tcp, 3888/tcp

Stop the images:

docker-compose down

Troubleshooting

docker-compose logs clickevent-generator|client|kafka|zookeeper|jobmanager|taskmanager-01|taskmanager-02|taskmanager-03
docker-compose images
docker container prune
docker system prune
docker image ls
docker image rm ID
docker run -i -t felipeogutierrez/tpch-dbgen /bin/bash

About

This project uses Apache Flink as a stream engine that consumes data from the File system or Kafka brokers and exposes metrics using Prometheus and Grafana, everything deployed on Kubernetes (minikube).

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages