Kubernetes

This section describes the Kubernetes deployment of PPC Robot - requirements for the cluster, node taints etc.

Kubernetes Version

PPC Robot should be run on Kubernetes 1.10 or newer. All resource types and APIs must be available in the 1.10.

Node Types and Labels

The following node labels are used for pod scheduling:

Label	Required by
nodes.ppc-robot.net/db	PostgreSQL database
nodes.ppc-robot.net/redis	Redis
nodes.ppc-robot.net/web	Web Interface
nodes.ppc-robot.net/scheduler	Scheduler
nodes.ppc-robot.net/worker-system	System tasks worker
nodes.ppc-robot.net/worker-cluster	Cluster management worker
nodes.ppc-robot.net/worker	Generic reports worker
nodes.ppc-robot.net/worker-url-check	URL Checkers worker

Values of the label should be set to true, however, only existence of such label is checked at the moment.

Deployment: Development Cluster

Before you begin, you have to add labels to corresponding nodes:

kubectl label nodes <node> nodes.ppc-robot.net/db=true
kubectl label nodes <node> nodes.ppc-robot.net/redis=true
kubectl label nodes <node> nodes.ppc-robot.net/web=true
kubectl label nodes <node> nodes.ppc-robot.net/scheduler=true
kubectl label nodes <node> nodes.ppc-robot.net/rate-limiter=true
kubectl label nodes <node> nodes.ppc-robot.net/worker-system=true
kubectl label nodes <node> nodes.ppc-robot.net/worker-cluster=true
kubectl label nodes <node> nodes.ppc-robot.net/worker=true
kubectl label nodes <node> nodes.ppc-robot.net/worker-url-check=true

To deploy PPC Robot in development, take the following steps (cd to the kubernetes directory first):

1. Create SSD volume type and prepare Persistent Volumes:

   kubectl apply -f development/ssd-storage-class.yml -f development/ssd-volumes.yml
   

2. Create necessary secrets and configmaps (update the configmaps/secrets if necessary):

   bash create-configmaps.sh
   bash create-secrets.sh
   

3. Deploy PostgreSQL and Redis:

   kubectl apply -f postgresql.yml -f redis.yml
   

4. Create database and user for ppc_robot:

   bash development/create-database.sh
   

5. Apply development configuration for PPC Robot:

   kubectl apply -f development/ppc-robot-config.yml
   

6. Deploy all containers and pods (requires the project to be installed on the machine):

   fab k8s_deploy_resources
   

Enabling access to the web UI: Add the following service that will bind the web UI to NodePort:

kubectl apply -f development/ppc-robot-web-nodeport.yml

Get the port on which the service was bound using the following command:

$ kubectl describe services/ppc-robot-web-nodeport
Name:                     ppc-robot-web-node-port
Namespace:                default
...
NodePort:                 <unset>  31830/TCP
...