Then, we will provision another three nodes as a fleet of Spot instances. Here, in the following screenshot, it shows the setup to request a fleet of three Spot instances. Choose the Request and Maintain option, then set the Target capacity to 3 instances:

We configure the setup script to install Docker, join the node to the cluster, and set up the network driver upon the instance creation. The setup must be put into the User data section of the fleet setup, as shown in the following screenshot:

Here's the script used in the User data section. Please replace $TOKEN with your worker's token, and $MANAGER_IP with one of your manager's private IP addresses:

#!/bin/bash
curl -sSL https://get.docker.com | sh
service docker start
usermod -aG docker ubuntu
docker swarm join --token $TOKEN $MANAGER_IP:2377
docker plugin install --grant-all-permissions weaveworks/net-plugin:2.1.3

Now, we wait until the fleet request is fulfilled.

If we get into the first manager, we could check the current nodes in the cluster with the docker node ls command. If everything is OK, we should have six nodes in the cluster:

$ docker node ls
ID          HOSTNAME          STATUS  AVAILABILITY   MANAGER STATUS
btul0hbd    ip-172-31-11-209  Ready   Active 
etm8veip    ip-172-31-8-157   Ready   Active 
iwl4pxnf *  ip-172-31-4-52    Ready   Active         Leader
rsqsflmv    ip-172-31-1-223   Ready   Active         Reachable
uxd36bok    ip-172-31-15-229  Ready   Active 
xn7fz2q1    ip-172-31-0-153   Ready   Active         Reachable

With this technique, we can easily scale the cluster by simply adjusting the number of Spot instances.