Put the monitoring system into practice. Don't try to make it perfect from the first attempt. You'll fail if you do. Iterate over dashboards. Start small, grow with time. If you are a bigger organization, let each team create their own set of dashboards and share what worked well, as well as what failed to provide enough value.
Monitoring is not a simple thing to do unless you want to spend all your time in front of a dashboard. The solution should be designed in such a way that you need only a glance to discover whether a part of the system requires your attention.
Now let us destroy everything we did. The next chapter will be a new subject with a new set of challenges and solutions:
docker-machine rm -f swarm-1 \
swarm-2 swarm-3 swarm-4 swarm-5