starting a consul cluster via fleet. 3 servers are started, and clients fill out the rest of the fleet. If you want to get specific, add `MachineMetadata` to specifically place where consul-servers will spend up

boot.sh

#!/bin/bash
# set -eo pipefail

export CONSUL_NUM_INSTANCES=${CONSUL_NUM_INSTANCES:-3}
echo " => stopping consul"
fleetctl stop -block-attempts=10 \
	$(eval echo "consul-server@{1..$CONSUL_NUM_INSTANCES}.service") \
	consul-agent.service

# FIXME: sometimes it takes awhile for systemd to stop things... 
# if we go right to destroy this can leave services running...
echo " => destroying consul"
fleetctl destroy \
	$(eval echo "consul-server@{1..$CONSUL_NUM_INSTANCES}.service") \
	consul-agent.service


echo " => sleeping and then submitting consul"
SCRIPT=`readlink -f "$0"`
SCRIPTPATH=`dirname "$SCRIPT"`
sleep 5
fleetctl load \
	$(eval echo "$SCRIPTPATH/consul-server@{1..$CONSUL_NUM_INSTANCES}.service") \
	"$SCRIPTPATH/consul-agent.service"


echo " => starting consul"
fleetctl start \
	$(eval echo "consul-server@{1..$CONSUL_NUM_INSTANCES}.service") \
	consul-agent.service

consul-agent.service

[Unit]
Description=Consul Agent
Requires=docker.service
Requires=fleet.service
After=docker.service
After=fleet.service
 
[Service]
TimeoutStartSec=10m
TimeoutStopSec=90s

Restart=on-failure
RestartSec=10s

# Get CoreOS environmental variables
EnvironmentFile=/etc/environment

ExecStartPre=-/usr/bin/docker kill %p
ExecStartPre=-/usr/bin/docker rm %p
ExecStartPre=/usr/bin/docker pull progrium/consul 
ExecStart=/bin/bash -c '\
	log_tag="%p"; \
	if $(systemctl list-units | grep consul-server 2>&1 >/dev/null); then \
		logger -t "$log_tag" "skipping host as it already has a consul server running"; \
		exit 0; \
	else \
		primary_ip=$(fleetctl list-units --fields=unit,sub,machine | grep -E "consul-server@1.service" | sed -r "s/.*\/(.+?)/\1/"); \
		logger -t "$log_tag" "starting consul client.  Connecting to $primary_ip"; \
		$(docker run --rm progrium/consul cmd:run ${COREOS_PRIVATE_IPV4}::$primary_ip::client --name %p); \
	fi; '

ExecStop=-/usr/bin/docker stop -t 30 %p
ExecStop=-/usr/bin/docker rm %p

[X-Fleet]
Global=true

consul-server@.service

[Unit]
Description=Consul Server
Requires=docker.service
Requires=fleet.service
After=docker.service
After=fleet.service
 
[Service]
TimeoutStartSec=10m
TimeoutStopSec=90s

Restart=always
RestartSec=10s

# Get CoreOS environmental variables
EnvironmentFile=/etc/environment

ExecStartPre=-/usr/bin/docker kill %p
ExecStartPre=-/usr/bin/docker rm %p
ExecStartPre=/usr/bin/docker pull progrium/consul 
ExecStart=/bin/bash -c '\
	log_tag="%p"; \
	if [ %i -eq 1 ]; then \
		logger -t "$log_tag" "starting initial consul server at ${COREOS_PRIVATE_IPV4}"; \
		$(docker run --rm -e EXPECT=3 progrium/consul cmd:run ${COREOS_PRIVATE_IPV4} --name %p); \
	else \
		primary_ip=$(fleetctl list-units --fields=unit,sub,machine | grep -E "%p@1.service" | sed -r "s/.*\/(.+?)/\1/"); \
		logger -t "$log_tag" "starting secondary consul server: $primary_ip"; \
		$(docker run --rm progrium/consul cmd:run ${COREOS_PRIVATE_IPV4}::$primary_ip --name %p); \
	fi; \
'

ExecStop=-/usr/bin/docker exec %p consul leave
ExecStop=-/usr/bin/docker stop -t 30 %p
ExecStop=-/usr/bin/docker rm %p
 
[X-Fleet]
Conflicts=%p@*.service