Unlike self-adaptation, self-healing is relatively easy to accomplish. As long as there are available resources, a scheduler will make sure that the specified number of replicas is always running. In our case, that scheduler is Docker Swarm (https://docs.docker.com/engine/swarm/).
Replicas can fail, they can be killed, and they can reside inside an unhealthy node. It does not really matter since Swarm will make sure that they are rescheduled when needed and (almost) always up-and-running. If all our services are scalable and we are running at least a few replicas of each, there will never be downtime. Self-healing processes inside Docker will make sure of that while our own self-adaptation processes aim to provide high-availability. The combination of the two is what makes the system almost fully autonomous and self-sufficient.
Problems begin piling...
