• ETCD data store has the following:
    • Nodes
    • PODs
    • Configs
    • Secrets
    • Accounts
    • Roles
    • Bindings
    • Other
  • Every piece of information you see when you run the kubectl get command is from the ETCD server.
  • Every additional change you make to your cluster - adding nodes, deploying pods or replica sets are updated in the ETCD server.
    • Only once the updates are updated in the ETCD server, is the change considered to be complete.
  • The practice test environments are deployed using the kubeadm tool. Later in the course when a cluster is set up, it is set up from scratch.
    • If you set up your cluster from scratch, you deploy ETCD by downloading the binaries yourself. This also includes configuring ETCD as a service on your Master Node. screenshot.png
  • You can pass many options such as certificates and these will be looked into further in the course.
    • There is a whole section of TLS certificates.
  • You can also configure ETCD as a cluster - part of high availability in Kubernetes.
  • The one option to note for now from the above screenshot is the --advertise-client-urls https://${INTERNAL_IP}:2379 \\.
    • This is the address where ETCD listens. This is on the IP of the server and is on port 2379. That is the default port which ETCD listens on.
      • This should be configured on the kube-api server, when it tries to reach the ETCD server.
  • If you cluster is using kubeadm, kubeadm will then deploy the ETCD server for you as a pod.
    • This will be in the kube-system namespace. screenshot.png
  • You can explore the ETCD database, by using the etcd control utility within the etcd-master pod.
  • To list all keys stored by Kubernetes, run the etcd control get command. 
    kubectl exec etcd-master -n kube-system etcdctl get / --prefix -keys-only
  • The output looks like the below: screenshot.png
  • Kubernetes stores data in a specific directory structure.
  • The root directory is the registry and under that you have various Kubernetes constructs. screenshot.png
  • ETCD in a High Availability Environment
    • You have multiple Master Nodes in the cluster. There are multiple ETCD nodes spread across the Master Nodes.
    • Make sure the etcd instances know about each other. Set the right parameters in the etcd service configuration under etcd.service. screenshot.png
  • The initial-cluster option is where you specify the different instances of the etcd service.

Updated: