I've been reading through: https://docs.docker.com/engine/swarm/
I'm interested in deploying a docker swarm using the latest version as of this writing (v18.09) on an rPi.
From what I understand, "nodes" in a swarm are distinct instances of the Docker Engine, and the level where RAFT consensus takes place. Given that fault tolerance is 0 for "node count"=1 or 2, I need to have at least 3 nodes in order to simulate fault-tolerance scenarios of a node going down.
All tutorials I can find only talk about each node as being on separate physical hosts, or in the case of an x86 machine with a hyper-visor, you run multiple instances of VM-guests via Virtualbox (or some other form of "full" virtualization) on the x86 host in order to simulate multiple nodes on a single physical host.
With an ARMx64 based linux physical host such as the rPi, running "full-fat" virtualization of a machine + guest OS in order to have separate Docker nodes would not be worthwhile since ARM doesn't have the hardware virtualization capabilities of an x86.
Is it possible to run more than one node / Docker Engine in parallel without having to do the "full" virtualization that an x86 host does more easily?