Docker Swarm is the Docker-native solution for deploying a cluster of Docker hosts. You can use it to quickly deploy a cluster of Docker hosts running either on your local machine or on supported cloud platforms.
Before Docker 1.12, setting up and deploying a cluster of Docker hosts required you to use an external key-value store like etcd or Consul for service discovery. With Docker 1.12, however, an external discovery service is no longer necessary, since Docker comes with an in-memory key-value store that works out of the box.
In this tutorial, you’ll learn how to deploy a cluster of Docker machines using the Swarm feature of Docker 1.12 on DigitalOcean. Each Docker node in the cluster will be running CentOS 7. While you can run a cluster made up of dozens, hundreds, or thousands of Docker hosts, the cluster we’ll be setting up in this tutorial will be made up of a manager node and two worker nodes, for a total of three cluster members. Once you complete this tutorial, you’ll be able to add more nodes to your cluster with ease.
For this tutorial, you’ll need:
We need to create several Docker hosts for our cluster. As a refresher, the following command provisions a single Dockerized host, where
$DOTOKEN is an environment variable that evaluates to your DigitalOcean API token:
- docker-machine create --driver digitalocean --digitalocean-image centos-7-0-x64 --digitalocean-access-token $DOTOKEN machine-name
Imagine having to do that to set up a cluster made up of at least three nodes, provisioning one host at a time.
We can automate the process of provisioning any number of Docker hosts using this command, combined with some simple Bash scripting. Execute this command on your local machine to create three Docker hosts, named
- for i in 1 2 3; do docker-machine create --driver digitalocean \
- --digitalocean-image centos-7-0-x64 \
- --digitalocean-access-token $DOTOKEN node-$i; done
After the command has completed successfully, you can verify that all the machines have been created by visiting your DigitalOcean dashboard, or by typing the following command:
- docker-machine ls
The output should be similar to the following, and it should serve as a quick reference for looking up the IP address of the nodes:
OutputNAME ACTIVE DRIVER STATE URL SWARM DOCKER ERRORS node-1 - digitalocean Running tcp://184.108.40.206:2376 v1.12.2 node-2 - digitalocean Running tcp://220.127.116.11:2376 v1.12.2 node-3 - digitalocean Running tcp://18.104.22.168:2376 v1.12.2
At this point, all three Dockerized hosts have been created, and you have each host’s IP address. They are also all running Docker 1.12.x, but are not yet part of a Docker cluster. In the next steps, we’ll configure the firewall rules that will make the nodes to function as members of a cluster, pick one of the nodes and make it the Docker Swarm manager, and configure the rest as Docker Swarm workers.
A cluster has to have at least one node that serves as a manager, though for a production setup, three managers are recommended. For this setup, let’s pick the first node and make it the Swarm manager. The other two nodes will be the worker nodes.
Certain network ports must be opened on the nodes that will be be part of a cluster for the cluster to function properly. That entails configuring the firewall to allow traffic through those ports. Because there are three different firewall applications that can be used to accomplish that task, the commands you need to execute on the nodes for each firewall application has been documented in a separate article. Follow this guide and configure the firewalls for each host. Open the proper ports on the manager, then repeat to open the ports on the two client nodes.
After you’ve completed this step, you can initialize the cluster manager.
We’ve decided that
node-1 will be our cluster manager, so log in to the node from your local machine:
- docker-machine ssh node-1
The command prompt will change to reflect the fact that you’re now logged into that particular node. To configure the node as the Swarm manager, type the following command:
- docker swarm init --advertise-addr node_ip_address
node_ip_address is the IP address of the node. You may get it from the output of
docker-machine ls or from your DigitalOcean dashboard.
You’ll see output that looks like the following:
OutputSwarm initialized: current node (2n4y8bwu6s7c1kwdryd945zv1) is now a manager. To add a worker to this swarm, run the following command: docker swarm join \ --token SWMTKN-1-57m6beujakwyx0t2j5mglpi9juf8czapp8kf2tj3gudkgjkoda-0eg0x08w7uvyxiacmmx0zv67o \ 22.214.171.124:2377 To add a manager to this swarm, run 'docker swarm join-token manager' and follow the instructions.
Within the output is the ID of the node, which is 2n4y8bwu6s7c1kwdryd945zv1 in this example, and the instructions on how to add the other nodes to the cluster.
So now you have a Docker Swarm with a manager configured. Let’s add the remaining nodes as workers.
To complete this step, you might want to open another terminal and leave the terminal tab or window you used to log into the Swarm manager alone for now.
First, connect to
node-2 from your local machine:
- docker-machine ssh node-2
Then execute this command, where
your_swarm_token is the token you received when you created the cluster in the previous step, and
manager_node_ip_address is the IP of the Swarm manager:
- docker swarm join \
- --token your_swarm_token \
After the command has been executed successfully, you’ll see this response:
OutputThis node joined a swarm as a worker.
Log out of
node-2, and then repeat this process with
node-3 to add it to your cluster.
You have now added two worker nodes to the cluster. If the firewall rules were configured correctly, you now have a functioning Docker Swarm, with all the nodes synchronized.
After the manager and worker nodes have been assigned to the cluster, all Docker Swarm management commands have to be executed on the manager nodes. So return to the terminal that you used to add the manager and type in this command to view all members of the cluster:
- docker node ls
The output should be similar to this:
OutputID HOSTNAME STATUS AVAILABILITY MANAGER STATUS 4smt8qechkvb8qp02nwo2oe6t * node-1 Ready Active Leader 5wzik3knakgj0c24jmtmiy5oq node-2 Ready Active 7f6ws6oladh94xnmbfdhanokm node-3 Ready Active
This output shows that we’re dealing with a 3-node Docker Swarm and its nodes — a manager and two workers. To view the other management commands that you can run on the manager node, type:
- docker node --help
For detailed information about the cluster, you may use the following command on the manager or workers (it’s a generic Docker command):
- docker info
The output should be of this sort, and should indicate the status of the cluster (active or pending), the number of nodes in the cluster, and whether the particular node is a manager or worker.
Output... Network: null host overlay bridge Swarm: active NodeID: 4smt8qechkvb8qp02nwo2oe6t Is Manager: true ClusterID: 31i554ti23njgxg28av52hv47 Managers: 1 Nodes: 3 Orchestration: Task History Retention Limit: 5 Raft: Snapshot Interval: 10000 Heartbeat Tick: 1 Election Tick: 3 Dispatcher: Heartbeat Period: 5 seconds CA Configuration: Expiry Duration: 3 months Node Address: 126.96.36.199 Runtimes: runc Default Runtime: runc Security Options: seccomp Kernel Version: 3.10.0-327.22.2.el7.x86_64 Operating System: CentOS Linux 7 (Core) OSType: linux ...
If you repeat the same command on the worker nodes, the Is Manager line should show
Tip: You can add or remove nodes from the cluster at any time. Additionally, a worker node can be promoted to a manager, and manager can be converted to a worker.
Now let’s get a service running on the cluster.
Now that you have a Docker Swarm up and running, let’s run a test container and see how the manager handles it. On a machine running Docker Engine 1.12 or newer, containers are deployed as Services using the
docker service command. And like the
docker node command, the
docker service command can only be executed on a manager node.
So let’s deploy a web server service using the official Nginx container image:
- docker service create -p 80:80 --name webserver nginx
In this command, we’re mapping port
80 in the Nginx container to port
80 on the cluster so that we can access the default Nginx page from anywhere.
To view which services are running on a cluster, type:
- docker service ls
The output should take this form. The REPLICAS column shows how many instances of the service are running:
OutputID NAME REPLICAS IMAGE COMMAND ch7xnvy8upho webserver 1/1 nginx
You can determine which nodes the services is running on by using
docker service ps followed by the service name.
- docker service ps webserver
The output should be similar to the following:
OutputID NAME IMAGE NODE DESIRED STATE CURRENT STATE ERROR 4pfbvs7sh5mugtzckw4czhcfm webserver.1 nginx node-1 Running Running 2 minutes ago
In this example, the
webserver service is running on
node-1. Since that’s a Web server running on the default ports, you can access it by pointing your browser to
http://node-1_ip_address. Give it a try. You’ll see Nginx’s default page.
With the magic of mesh networking, a service running on a node can be accessed on any other node of the cluster. For example, this Nginx service can also be accessed by pointing your browser to the IP address of any node in the cluster, not just the one it is running on. Give it a try.
Another feature of Docker Swarm is the ability to scale a service, that is, spin up additional instances of a service. Assume that we want to scale the
webserver service that we started earlier to five instances. To do so, we just type the following command and the system will create four more instances:
- docker service scale webserver=5
And the output of
docker service ps will show on which nodes the new instances were started on:
OutputID NAME IMAGE NODE DESIRED STATE CURRENT STATE ERROR 4pfbvs7sh5mugtzckw4czhcfm webserver.1 nginx node-1 Running Running 3 minutes ago 1er2rbrnj6ltanoe47mb653wf webserver.2 nginx node-3 Running Running 14 seconds ago evassgyvruh256ebv5pj3bqcz webserver.3 nginx node-3 Running Running 14 seconds ago d453agrdpgng47klbl6yfjnka webserver.4 nginx node-1 Running Running 18 seconds ago 2hsdevx178rg15gqxhzrsnmg6 webserver.5 nginx node-2 Running Running 14 seconds ago
That shows that two of the four new instances were started on
node-3, one was started on
node-1 and the other started on
Finally, if a service goes down, it’s automatically restarted on the same node or on a different node, if the original node is no longer available.
You’ve seen how easy it is to set up a Docker Swarm using Docker Engine 1.12 and the new Swarm mode. You’ve also seen how to perform a few management tasks on the cluster. But there’s still more. To view the available Docker Swarm commands, execute the following command on your Swarm manager.
- docker swarm --help
If you’ve enjoyed this tutorial and our broader community, consider checking out our DigitalOcean products which can also help you achieve your development goals.