It’s a cluster comprised of a group of physical and virtual machines running the Docker Engine, these are called nodes.

One of the nodes is elected to be the Leader by using the Raft consensus algorithm. He make all the orchestration decisions for the swarm. In case of failure, a new leader it’s automatically elected.

Manager nodes assign tasks to Worker nodes. They can do some other managerial tasks. You want to have an odd number of managers, for reliability purposes. Having to many managers can degrade performance, it’s recommended to have at most 7.

Overlay networks manage communications among the Docker nodes participating to the Swarm. Can also be used for service-to-service communication.

Ingress network is a special overlay network that facilitates load balancing among the Swarm nodes.

docker_gwbridge is a network connecting the containers running on a Docker daemon’s host. Also this network connects the overlay networks to a individual Docker daemon’s physical network