Sample Flask Project with standard DevOps structure wrapped as a Docker container and deployed to K8S on IBM Cloud

• To run the project locally: (Python 3.0+). Add any libraries you use inside requirements.txt

    pip install -r requirements.txt
    python app.py
    curl localhost:5000/noop (returns nothing)
    curl -X POST localhost:5000/jobs
  

• Build and push the project to Image Registry (in this case IBM Cloud)

    docker build -t *|name|* .
    docker push *|name|* *|remote_host|*
  

• Update the app, name fields in all .yaml files along with the container and remote host names

• Now, setup access to your K8S Cluster from your cluster. We are using kubectl in this case

    ibmcloud login -a cloud.ibm.com -r us-south -g default -sso
    ibmcloud ks cluster-config --cluster <cluster-name>
    export ...
    kubectl get nodes
  

• Create a Persistent Volume Claim. Make sure your account has access to IBM Storage and your pvc.yaml file has been configured. PVC creation will take a couple of minutes. Wait until it finishes.

    kubectl apply -f k8s/storage/pvc.yaml.
  

• Apply deployment and service files to K8S. Make sure all name, app fields are changed according to your preference.

    kubectl apply -f k8s/basic/deployment.yaml
    kubectl apply -f k8s/basic/service.yml
  

• To run the project, use External IP of one of the nodes and a service PORT that you defined in your service.yaml file. Change Port number if it collides with another service.

    kubectl describe nodes
    curl *|external_ip|*:*|port|*/noop
  

You can setup the ingress to work with your service. Then, you can access your service using the hostname provided in your ingress.yaml file. e.g curl example.us-south.containers.appdomain.cloud/noop

Structure of the project

Notification Service uses synchronous job id schema in which a request that has been received will be assigned a job id and started off as a separate thread. Job ID will be returned to the client synchronously (immediately). Client can check the status of the job, by invoking '/jobs/|job_id|'. Within the system, job gets tracked using a status.txt file which gets updated every time a new phase of the execution kicks in.

Structure of the Cloud Deployment

pvc.yaml - Stands for Persistent Volume Claim. This uses the admin account to provision storage on IBM Cloud Storage with NFS Provisioning. This way, there is no need to manually allocate new PVs. Once the storage is provisioned, PVC also authorizes itself to access the storage, which is transitive to the Deployments that implement this PVC.

deployment.yaml - Creates as many pods as specified in the file, as well as does attaching PVC and setting up a mount path. Remote host of the Docker image that needs to be loaded also has to be specified in this file, along with which port will expose your server locally.

service.yml - Exposes your deployment/pods to the outside world. 5 digit port number can be specified as the access point to a specific deployment under the field NodePort. There is also a field with the name 'Port' which refers the local port that serves inside the container. i.e 5000 for Flask.

Invocations and Endpoints

1. Noop

• Endpoint:                               'hostname':'port'/noop
• Type:                                   GET
• Return:                                 204

2. Create Jobs

• Endpoint:                               'hostname':'port'/jobs
• Type:                                   POST
• Return:                                 200, JSON with status

3. Check Notification Job ID

• Endpoint:                               'hostname':'port'/jobs/
• Type:                                   GET
• Return:                                 200, JSON with status

@author Created by Jam Mirzakhalov, 2019