A CloudFormation template to build a CoreOS-based Docker cluster on AWS using Swarm.
The CloudFormation template takes the following parameters:
| Parameter | Description |
|---|---|
| InstanceType | EC2 HVM instance type (t2.micro, m3.medium, etc.) for the Swarm master and node hosts. |
| ClusterSize | Number of nodes in the Swarm cluster (3-12). |
| DiscoveryURL | A unique etcd cluster discovery URL. Grab a new token from https://discovery.etcd.io/new?size=4 |
| AdvertisedIPAddress | Use 'private' if your etcd cluster is within one region or 'public' if it spans regions or cloud providers. |
| AllowSSHFrom | The net block (CIDR) from which you can use SSH and docker to communicate with the Swarm master. |
| KeyName | The name of an EC2 Key Pair to allow SSH access to the Swarm master. |
| VpcAvailabilityZones | Comma-delimited list of three VPC availability zones in which to create subnets. |
The template builds a new VPC with 3 subnets (in 3 availability zones), a single Swarm master and a cluster of between 3 and 12 nodes, using the standard AWS CoreOS AMI.
All hosts form part of an initial etcd cluster for node discovery during bootstrapping. Therefore the size option on the discovery.etcd.io URL must be one more than the size of the CoreOS cluster.
Hosts are evenly distributed across the 3 availability zones and are created within an auto-scaling group which can be manually adjusted to alter the Swarm cluster size post-launch.
A 'docker-swarm' container is run on each host (the Swarm master and the nodes). All hosts listen on port 4243, leaving the standard docker port (2375) free for use by the 'docker-swarm' container on the master.
Separate 'master' and 'node' security groups control access between the nodes. The template builds the Swarm master first, then the auto-scaling group for the nodes (each of which needs to register with the master via its 'docker-swarm' container).
Note that each host also runs the Fleet service, which can be used as an alternative cluster manager.
| Output | Description |
|---|---|
| MasterDockerPs | Command to run a 'docker ps' on the Swarm master |
| MasterPublicIP | Public IP for the cluster master |
| MasterPrivateIP | Private IP for the cluster master |
Try
docker -H tcp://_MasterPublicIP_:2375 info
docker -H tcp://_MasterPublicIP_:2375 run docker/whalesay cowsay 'It works!'
to see if the cluster is working.
For debugging, you should be able to ssh to the master, using the username 'core' and your private key. You can ssh on from there to each of the nodes. Other than that, it shouldn't really be necessary to log in to any of the hosts.
- Persistent storge - no, there isn't any currently!
- Hybrid swarms - building a cluster across vCloud Director / vCloud Air and AWS. See https://github.com/andypowe11/vCloudDirector-Docker-Swarm for a vca-cli script to build the equivalent cluster on vCloud Director or vCloud Air.