Build a Swarm cluster for production
Estimated reading time: 10 minutes
You are viewing docs for legacy standalone Swarm. These topics describe standalone Docker Swarm. In Docker 1.12 and higher, Swarm mode is integrated with Docker Engine. Most users should use integrated Swarm mode — a good place to start is Getting started with swarm mode, Swarm mode CLI commands, and the advanced Learn Docker sample app. Standalone Docker Swarm is not integrated into the Docker Engine API and CLI commands.
This page teaches you to deploy a high-availability Docker Swarm cluster. Although the example installation uses the Amazon Web Services (AWS) platform, you can deploy an equivalent Docker Swarm cluster on many other platforms. In this example, you do the following:
- Verify you have the prerequisites
- Establish basic network security
- Create your nodes
- Install Engine on each node
- Configure a discovery backend
- Create Swarm cluster
- Communicate with the Swarm
- Test the high-availability Swarm managers
- Additional Resources
For a gentler introduction to Swarm, try the Evaluate Swarm in a sandbox page.
- An Amazon Web Services (AWS) account
- Familiarity with AWS features and tools, such as:
- Elastic Cloud (EC2) Dashboard
- Virtual Private Cloud (VPC) Dashboard
- VPC Security groups
- Connecting to an EC2 instance using SSH
Step 1. Add network security rules
AWS uses a “security group” to allow specific types of network traffic on your VPC network. The default security group’s initial set of rules deny all inbound traffic, allow all outbound traffic, and allow all traffic between instances.
You’re going to add a couple of rules to allow inbound SSH connections and inbound container images. This set of rules somewhat protects the Engine, Swarm, and Consul ports. For a production environment, you would apply more restrictive security measures. Do not leave Docker Engine ports unprotected.
From your AWS home console, do the following:
Click VPC - Isolated Cloud Resources.
The VPC Dashboard opens.
Navigate to Security Groups.
Select the default security group that’s associated with your default VPC.
Add the following two rules.
Type Protocol Port Range Source SSH TCP 22 0.0.0.0/0 HTTP TCP 80 0.0.0.0/0
The SSH connection allows you to connect to the host while the HTTP is for container images.
Step 2. Create your instances
In this step, you create five Linux hosts that are part of your default security group. When complete, the example deployment contains three types of nodes:
|Swarm primary and secondary managers||
To create the instances do the following:
Open the EC2 Dashboard and launch five EC2 instances, one at a time.
During Step 1: Choose an Amazon Machine Image (AMI), pick the Amazon Linux AMI.
During Step 5: Tag Instance, under Value, give each instance one of these names:
During Step 6: Configure Security Group, choose Select an existing security group and pick the “default” security group.
Review and launch your instances.
Step 3. Install Engine on each node
In this step, you install Docker Engine on each node. By installing Engine, you enable the Swarm manager to address the nodes via the Engine CLI and API.
SSH to each node in turn and do the following.
Update the yum packages.
Keep an eye out for the “y/n/abort” prompt:
$ sudo yum update
Run the installation script.
$ curl -sSL https://get.docker.com/ | sh
"-H tcp://0.0.0.0:2375 -H unix:///var/run/docker.sock"to the
$ sudo /etc/init.d/docker start
Verify that Docker Engine is installed correctly by running a container with the
$ sudo docker run hello-world
The output should display a “Hello World” message and other text without any error messages.
$ sudo usermod -aG docker ec2-user
If entering a
dockercommand produces a message asking whether docker is available on this host, it may be because the user doesn’t have root privileges. If so, use
sudoor give the user root privileges.
For this example, don’t create an AMI image from one of your instances running Docker Engine and then re-use it to create the other instances. Doing so will produce errors.
If your host cannot reach Docker Hub, the
docker runcommands that pull container images may fail. In that case, check that your VPC is associated with a security group with a rule that allows inbound traffic (e.g., HTTP/TCP/80/0.0.0.0/0). Also Check the Docker Hub status page for service availability.
Step 4. Set up a discovery backend
Here, you’re going to create a minimalist discovery backend. The Swarm managers and nodes use this backend to authenticate themselves as members of the cluster. The Swarm managers also use this information to identify which nodes are available to run containers.
To keep things simple, you are going to run a single consul daemon on the same host as one of the Swarm managers.
Use SSH to connect to the
From the output, copy the
eth0IP address from
To set up a discovery backend, use the following command, replacing
<consul0_ip>with the IP address from the previous command:
$ docker run -d -p 8500:8500 --name=consul progrium/consul -server -bootstrap -advertise=<consul0_ip>
From the output, verify that a consul container is running. Then, disconnect from the
Your Consul node is up and running, providing your cluster with a discovery backend. To increase its reliability, you can create a high-availability cluster using a trio of consul nodes using the link mentioned at the end of this page. (Before creating a cluster of consul nodes, update the VPC security group with rules to allow inbound traffic on the required port numbers.)
Step 5. Create Swarm cluster
After creating the discovery backend, you can create the Swarm managers. In this step, you are going to create two Swarm managers in a high-availability configuration. The first manager you run becomes the Swarm’s primary manager. Some documentation still refers to a primary manager as a “master”, but that term has been superseded. The second manager you run serves as a replica. If the primary manager becomes unavailable, the cluster elects the replica as the primary manager.
Use SSH to connect to the
manager0instance and use
ifconfigto get its IP address.
To create the primary manager in a high-availability Swarm cluster, use the following syntax:
$ docker run -d -p 4000:4000 swarm manage -H :4000 --replication --advertise <manager0_ip>:4000 consul://<consul0_ip>:8500
<consul0_ip>with the IP address from the previous command, for example:
$ docker run -d -p 4000:4000 swarm manage -H :4000 --replication --advertise 172.30.0.125:4000 consul://172.30.0.161:8500
From the output, verify that a Swarm cluster container is running. Then, disconnect from the
Connect to the
manager1node and use
ifconfigto get its IP address.
Start the secondary Swarm manager using following command.
<manager1_ip>with the IP address from the previous command, for example:
$ docker run -d -p 4000:4000 swarm manage -H :4000 --replication --advertise <manager1_ip>:4000 consul://172.30.0.161:8500
docker psto verify that a Swarm container is running. Then disconnect from the
node1in turn and join them to the cluster.
a. Get the node IP addresses with the
b. Start a Swarm container each using the following syntax:
docker run -d swarm join --advertise=<node_ip>:2375 consul://<consul0_ip>:8500
$ docker run -d swarm join --advertise=172.30.0.69:2375 consul://172.30.0.161:8500
docker psto verify that the Swarm cluster container started from the previous command is running.
Your small Swarm cluster is up and running on multiple hosts, providing you with a high-availability virtual Docker Engine. To increase its reliability and capacity, you can add more Swarm managers, nodes, and a high-availability discovery backend.
Step 6. Communicate with the Swarm
You can communicate with the Swarm to get information about the managers and
nodes using the Swarm API, which is nearly the same as the standard Docker API.
In this example, you use SSL to connect to
consul0 host again.
Then, you address commands to the Swarm manager.
Get information about the manager and nodes in the cluster:
$ docker -H :4000 info
The output gives the manager’s role as primary (
Role: primary) and information about each of the nodes.
Run an application on the Swarm:
$ docker -H :4000 run hello-world
Check which Swarm node ran the application:
$ docker -H :4000 ps
Step 7. Test Swarm failover
To see the replica instance take over, you’re going to shut down the primary manager. Doing so kicks off an election, and the replica becomes the primary manager. When you start the manager you shut down earlier, it becomes the replica.
SSH connection to the
Get the container id or name of the
$ docker ps
Shut down the primary manager, replacing
<id_name>with the container’s id or name (for example, “8862717fe6d3” or “trusting_lamarr”).
docker rm -f <id_name>
Start the Swarm manager. For example:
$ docker run -d -p 4000:4000 swarm manage -H :4000 --replication --advertise 172.30.0.161:4000 consul://172.30.0.161:8500
Review the Engine’s daemon logs the logs, replacing
<id_name>with the new container’s id or name:
$ sudo docker logs <id_name>
The output shows will show two entries like these ones:
time="2016-02-02T02:12:32Z" level=info msg="Leader Election: Cluster leadership lost" time="2016-02-02T02:12:32Z" level=info msg="New leader elected: 172.30.0.160:4000"
To get information about the manager and nodes in the cluster, enter:
$ docker -H :4000 info
You can connect to the
manager1 node and run the
They will display corresponding entries for the change in leadership.