The team at Prometheus describes the product as:

Prometheus's main features are:

There is a great presentation on https://prometheus.io/docs/introduction/overview/ that we will not repeat here. The top feature of Prometheus is, in our opinion, the ease of installation and usage. Prometheus itself consists of just a single binary built from Go code, plus a configuration file.

Things are probably going to change very soon, so we just sketch a way to set up a monitoring system for Swarm, tried on Docker version 1.12.3.

First, we create a new overlay network to not interfere with the ingress or spark networks, called monitoring:

aws-101$ docker network create --driver overlay monitoring

Then, we start a cAdvisor service in mode global, meaning that a cAdvisor container will run on each Swarm node. We mount some system paths inside the container so that they can be accessed by cAdvisor:

aws-101$ docker service create 
  --mode global 
  --name cadvisor 
  --network monitoring 
  --mount type=bind,src=/var/lib/docker/,dst=/var/lib/docker 
  --mount type=bind,src=/,dst=/rootfs 
  --mount type=bind,src=/var/run,dst=/var/run 
  --publish 8080 
  google/cadvisor

Then we use basi/prometheus-swarm to set up Prometheus:

aws-101$ docker service create 
  --name prometheus 
  --network monitoring 
  --replicas 1 
  --publish 9090:9090 
  prom/prometheus-swarm

And we add the node-exporter service (again global, must run on each node):

aws-101$ docker service create 
  --mode global 
  --name node-exporter 
  --network monitoring 
  --publish 9100 
  prom/node-exporter

Finally, we start Grafana with one replica:

aws-101$ docker service create 
  --name grafana 
  --network monitoring 
  --publish 3000:3000 
  --replicas 1 
  -e "GF_SECURITY_ADMIN_PASSWORD=password" 
  -e "PROMETHEUS_ENDPOINT=http://prometheus:9090" 
  grafana/grafana

When Grafana is available, to get impressive graphs of the Swarm health, we login with these credentials on the node where Grafana runs, port 3000:

"admin":"password"

As admins, we click on the Grafana logo, go to Data Sources, and add Prometheus:

Some options will appear, but the mapping is already present, so it's sufficient to Save & Test:

Now we can go back to the Dashboard and click on Prometheus, so we will be presented the Grafana main panel:

Once again, we took advantage of what the open source community released, and glued different opinionated technologies with just some simple commands, to get the desired result. Monitoring Docker Swarm and its applications is a field of research that is completely open now, so we can expect an amazing evolution there too.