CRU-D stands for "CReate, Update minus the Delete" operations. It is a quick-mini project to demonstrate common hypothetical create & update operations on such online job search site. The project was deployed in a single node Kubernate cluster.
Hardware & Sofware specification:
- Windows OS
- Memory 4Gb
- JDK 8
- Intellij IDEA
- Docker Dekstop
- Minikube
- Kubectl
Kubernates (K8S) - single node cluster
- 1 node
- 2 replica pots for the app
- 1 pot for the Database
Based on Windows & Docker Dekstop
Download the Docker Dekstop, Minikube, kubectl. See the internet to read how to set them up on different OS.
- Start the Minikube container (it will pull the Minikube (K8S engine). The image is gcr.io/k8s-minikube/kicbase)
minikube start --driver=docker
- Sync the terminal with the Kubernate VM context.
*Note : This is important step. This is to configure Docker CLI to access Docker from Minikube VM so that later we dont run containers within our local host machine but inside minikube VM docker daemon.
CMD
@FOR /f "tokens=*" %i IN ('minikube -p minikube docker-env --shell cmd') DO @%i
or with Powershell
& minikube -p minikube docker-env --shell powershell | Invoke-Expression
or with bash/zsh
eval $(minikube -p <profile> docker-env)
- Pull the project image
docker pull fevly/job-portal-crud-kube:1.0
and pull the MySQL 5.7 image
docker pull MySQL 5.7
- From the terminal get into the project root directory then apply the database deployment object and service defined in database-deployment.yaml
For e.g.
D:\Fevly\job-portal-crud-main>kubectl apply -f database-deployment.yaml
You can log the pods with :
kubectl logs -f <the_pod_name>
- Close the terminal. Open it and go to root project directory again and apply the app deployment object & service (jobportal-deployment.yaml). There will be 2 replica pods spawn.
kubectl apply -f jobportal-deployment.yaml
log the app to make sure the container runs properly
- Start the app service with port-forwarding mechanism so the app can interact outside the cluster
See the list of services deployed along with their internal cluster IP and ports
kubectl get svc
then start the app service with forwarding the port
kubectl port-forward service/<name_of_service> <outside_target_port>:<app_port>
For e.g:
kubectl port-forward service/job-portal-crud-svc 7080:8080
So in this case the services availabe on http://localhost:7080/
- Use Postman to access the APIs on http://localhost:7080/
Untuk coba URL-URL (endpoint) APInya bisa di lihat di dalam folder \job-portal-crud-main\src\main\resources. Filenya bernama DATA TEST.txt
Projectnya hanya dibangun dalam (kurang lebih) satu harian jadi tidak cukup banyak data dummy yang sempat dibuat
Based on Windows OS
- Enable WSL (Windows Subsystem for Linux)
- Start the Docker Deskstop.
- Enter the project root directory with the Terminal
- Start the Minikube
minikube start --driver=docker
It will pull the Kubernates/K8S engine image (gcr.io/k8s-minikube/kicbase) into the local Docker registry
- Sync the terminal with the Kubernate context.
Note : This is an important step. This is to configure Docker CLI to access Docker from Minikube VM so that later we dont run containers within our local host machine but inside minikube VM docker deamon.
CMD
@FOR /f "tokens=*" %i IN ('minikube -p minikube docker-env --shell cmd') DO @%i
Powershell
& minikube -p minikube docker-env --shell powershell | Invoke-Expression
- Pull the mysql 5.7 image into internal Kubernates's VM docker engine
- Note : The database-deployment.yaml mentioned previously is supposed to automatically pull this images into the local registry. However in my case, pull operation was for some unknown reasons often skipped. Hence, manually pulling needed.*
- Build the project image into the internal Kubernates's VM docker daemon
docker build -t job-portal-crud-kube:1.0 .
- Apply the database deployment object & service (database-deployment.yaml)
kubectl apply -f database-deployment.yaml
The pod should be now in running state, check it with kubectl get pods. If there are any errors then try to debug it with :
kubectl logs <that_pod_name>
or
kubectl describe pods <that_pod_name>
- Apply the app deployment object & service (jobportal-deployment.yaml). There will be 2 replica pods spawn.
kubectl apply -f jobportal-deployment.yaml
Note that If you (only if you) log the app pods and get exceptions such as "Unknown database host" or any issues related to the database connection then there is an issue with the DNS resolution. The app doesn't connect to the DB through some exposed services instead, through the DNS described in the database-deployment.yaml . This kind of problems often occurs when running Docker in Windows OS which resides inside the WSL kernel. It's known that debugging DNS issue in a local K8S cluster is quite tricky. It took me hours to debug. But here are some suggestions:
- Restart the Kubernate
corednswith :
kubectl -n kube-system rollout restart deployment coredns
-
If you already have configured some PersistentVolume(provisioned storage in the cluster ) then delete that. Or delete the entire Minikube container and start over
-
Delete all running pods with
kubectl delete all --all --all-namespaces
Then stop & start the minikube
minikube stop
minikube start --driver=docker
then execute these steps in order:
@FOR /f "tokens=*" %i IN ('minikube -p minikube docker-env --shell cmd') DO @%i
kubectl apply -f database-deployment.yaml
close the terminal, open new one then do
kubectl apply -f jobportal-deployment.yaml
The order in the execution steps above may sounds like a weird anomaly but it works in my case. Could it be because Cofiguring Docker CLI to access Docker from Minikube VM is somehow breaking the DNS resolution within the cluster? And why it works straightforwardly on other OS such as macOS? Not sure.
- It's done now expose the app service then access the APIs.