Short and sweet one on the Kubelet today.
Kubelet is like a captain on a ship in the ship analogy - it asks for the paperwork necessary to become part of the cluster. It is sole point of contact from the master/control plane ship, it loads and unloads containers on the ship as instructed by the Scheduler, and it sends back reports on the status of the ship as well as the containers on them.
The Kubelet registers the worker node with a Kubernetes cluster. When it receives instructions to load a pod on the node with a container, it requests the container runtime engine to pull and run an image. The Kubelet will then continue to monitor the state of the pod and containers in it, and then reports this back to the Kube API server.
I personally think of the Kubelet as the manager of the node and what's going on on it - the fact that it's the one doing status reports to the API server encapsulates what I think of when I think of the Kubelet, especially as it needs to know the state and status of the pods in this way. Also it's the one instructing the container runtime.
Even with the Kube Admin Tool, the Kubelet needs to be manually installed on the worker nodes - download the installer, extract it, and run it as a service.
You can view the running Kubelet process and options by listeing the process on the worker node and search for kubelet with ps aux | grep kubelet
Sweet and simple!
- Catt
Top comments (0)