- How these work together, in order to create a pod.
- Time Sequence Diagram –> Shows the sequential order of actions.
- How does a pod get scheduled on to a node?
- Kube config file gives permissions to
kubectl to communicate with the Kubernetes cluster.
kubectl command sends information to the kube-api server. That then saves the new deployment spec in etcd
- While this is occurring, the Kube Controller Manager is checking the Kube API Server, to see if there has been any changes since its last loop:
- Since there is a new deployment, the Kube Scheduler checks the Kube API Server, to see if there are any new pods that have not been assigned a node.
- Kube API Server tells the Scheduler that there is a new pod, that has not been placed on a specific node.
- Therefore, the Dcheduler chooses a node for the pod and sends the information back to the Kube API Server:
- The API Server saves the state of the cluster in
etcd:
- Regarding the Worker Node, the kubelet checks the API Server, to see if there are any new pods that it has been assigned.
- Kube API Server then sends the pod spec for the new pod to the Kubelet. This pulls the image and then creates the container, using the container runtime:
- Once the pod has been created, the Kubelet sends the pod status (healthy or unhealthy), back to the API server:
- This then saves the state to
etcd:
- Kube API Server processes thousands of requests to keep the server running.
- Useful talk called “Beyond Block Diagrams” on this subject.