InfraKit

InfraKit is a toolkit for infrastructure orchestration. With an emphasis on immutable infrastructure, it breaks down infrastructure automation and management processes into small, pluggable components. These components work together to actively ensure the infrastructure state matches the user's specifications. InfraKit therefore provides infrastructure support for higher-level container orchestration systems and can make your infrastructure self-managing and self-healing.

To get started, try the tutorial, or check out the video below:

InfraKit + LinuxKit POC

In this video, InfraKit was used to build a custom linux operating system (based on linuxkit). We then deployed a cluster of virtual machine instances on a local Mac laptop using the Mac Xhyve hypervisor (HyperKit). A cluster of 3 servers booted up in seconds. Later, after the custom OS image has been updated with a new public key, InfraKit detects the change and orchestrates a rolling update of the nodes. We then deploy the same OS image to a bare-metal ARM server running on Packet.net, where the server uses custom ipxe boot directly from the localhost. It demonstrates some of the key concepts and components in InfraKit and shows how InfraKit can be used to implement an integrated workflow from custom OS image creation to cluster deployment and Day N management. The entire demo is published as a playbook, and you can create your own playbooks too.

Use Cases

InfraKit is designed to automate setup and management of infrastructure in support of distributed systems and higher-level container orchestration systems. Some of the use cases we are working on include:

• Bootstrap / installation of container orchestration systems like Docker Swarm and Kubernetes
• Cluster autoscaler that can work across a variety of platforms from public clouds (like AWS autoscaling groups) to bare-metal hosts.
• GPU cluster provisioning
• Integration with LinuxKit for building and deploying immutable infrastructure from declarative specifications of the entire stack: from infrastructure resources to os / kernel and applications.
• Day-N management and automation of infrastructure - from provisioning to rolling updates and capacity scaling.

InfraKit has a modular architecture with a set of interfaces which define the interactions of these 'plugin objects'. Plugins are active daemons that cooperate with one another to ensure the infrastructure state matches your specifications.

Plugins

InfraKit makes extensive use of Plugins to manage arbitrary systems in diverse environments, which can be composed to meet different needs. See the plugins documentation for more technical details.

Here is a list of plugins:

Core Implementations

plugin	type	description
group	group	core group controller for rolling updates, scale group, etc.
swarm	flavor	runs Docker in Swarm mode
kubernetes	flavor	bootstraps a single master kubernetes cluster
combo	flavor	combine multiple flavor plugins
vanilla	flavor	manual specification of instance fields
aws	instance	creates Amazon EC2 instances and other resource types
digitalocean	instance	creates DigitalOcean droplets
docker	instance	provisions container via Docker
google	instance	Google Cloud Platform compute instances
file

Community Implementations

plugin	type	description
HewlettPackard/infrakit-instance-oneview	instance	bare-metal server provisioning via HP-OneView
IBM Cloud	instance	Provisions instances on IBM Cloud via terraform
AliyunContainerService/infrakit.aliyun	instance	Provisions instances on Alibaba Cloud
1and1/infrakit-instance-oneandone	instance	Provisions instances on 1&1 Cloud Server
sacloud/infrakit-instance-sakuracloud	instance	Provisions instances on Sakura Cloud

Have a Plugin you'd like to share? Submit a Pull Request to add yourself to the list!

Building

Your Environment

Make sure you check out the project following a convention for building Go projects. For example,

# Install Go - https://golang.org/dl/
# Assuming your go compiler is in /usr/local/go
export PATH=/usr/local/go/bin:$PATH

# Your dev environment
mkdir -p ~/go
export GOPATH=!$
export PATH=$GOPATH/bin:$PATH

mkdir -p ~/go/src/github.com/docker
cd !$
git clone git@github.com:docker/infrakit.git
cd infrakit

We recommended go version 1.9 or greater for all platforms.

Also install a few build tools:

make get-tools

Running tests

$ make ci

Binaries

$ make binaries

Executables will be placed in the ./build directory. There is only one executable infrakit which can be used as CLI and as server, based on the CLI verbs and flags.

Design

Configuration

InfraKit uses JSON for configuration because it is composable and a widely accepted format for many infrastructure SDKs and tools. Since the system is highly component-driven, our JSON format follows simple patterns to support the composition of components.

A common pattern for a JSON object looks like this:

{
   "SomeKey": "ValueForTheKey",
   "Properties": {
   }
}

There is only one Properties field in this JSON and its value is a JSON object. The opaque JSON value for Properties is decoded via the Go Spec struct defined within the package of the plugin -- for example -- vanilla.Spec.

The JSON above is a value, but the type of the value belongs outside the structure. For example, the default Group Spec is composed of an Instance plugin, a Flavor plugin, and an Allocation:

{
  "ID": "name-of-the-group",
  "Properties": {
    "Allocation": {
    },
    "Instance": {
      "Plugin": "name-of-the-instance-plugin",
      "Properties": {
      }
    },
    "Flavor": {
      "Plugin": "name-of-the-flavor-plugin",
      "Properties": {
      }
    }
  }
}

The group's Spec has Instance and Flavor fields which are used to indicate the type, and the value of the fields follow the pattern of <some_key> and Properties as shown above.

The Allocation determines how the Group is managed. Allocation has two properties:

• Size: an integer for the number of instances to maintain in the Group
• LogicalIDs: a list of string identifiers, one will be associated with each Instance

Exactly one of these fields must be set, which defines whether the Group is treated as 'cattle' (Size) or 'pets' (LogicalIDs). It is up to the Instance and Flavor plugins to determine how to use LogicalID values.

As an example, if you wanted to manage a Group of NGINX servers, you could write a custom Group plugin for ultimate customization. The most concise configuration looks something like this:

{
  "ID": "nginx",
  "Plugin": "my-nginx-group-plugin",
  "Properties": {
    "port": 8080
  }
}

However, you would likely prefer to use the default Group plugin and implement a Flavor plugin to focus on application-specific behavior. This gives you immediate support for any infrastructure that has an Instance plugin. Your resulting configuration might look something like this:

{
  "ID": "nginx",
  "Plugin": "group",
  "Properties": {
    "Allocation": {
      "Size": 10
    },
    "Instance": {
      "Plugin": "aws",
      "Properties": {
        "region": "us-west-2",
        "ami": "ami-123456"
      }
    },
    "Flavor": {
      "Plugin": "nginx",
      "Properties": {
        "port": 8080
      }
    }
  }
}

Once the configuration is ready, you will tell a Group plugin to

• watch it
• update it
• destroy it

Watching the group as specified in the configuration means that the Group plugin will create the instances if they don't already exist. New instances will be created if for any reason existing instances have disappeared such that the state doesn't match your specifications.

Updating the group tells the Group plugin that your configuration may have changed. It will then determine the changes necessary to ensure the state of the infrastructure matches the new specification.

Docs

Additional documentation can be found here.

Reporting security issues

The maintainers take security seriously. If you discover a security issue, please bring it to their attention right away!

Please DO NOT file a public issue, instead send your report privately to [security@docker.com](mailto:security@docker.com