The etcd database is an important resource because all cluster data is stored there, including all Kubernetes objects such as cluster-scoped resources, and deployment and pod information. If there is a disaster that causes the primary node to go down, the Kubernetes cluster suffers unavailability, and no operations can be scheduled or managed. The DevOps team must have a procedure to follow regularly to protect the etcd database. The team also requires a procedure to restore the database from the protection system to bring the system back online and work to the last-known backup timestamp. 

You can deploy the etcd database as a pod in the primary node and also deploy it externally to enable resiliency and security. Also, Kubernetes deployments can be of different types such as bare-metal Linux nodes, or they can be deployed as a virtual machine.

Figure 1.   Deployment topologies for etcd