InfraKit

<!-- [](https://circleci.com/gh/docker/infrakit) -->

<!-- [](https://codecov.io/github/docker/infrakit?branch=master) -->

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 | instance | useful for development and testing | | hyperkit | instance | creates HyperKit VMs on Mac OSX | | libvirt | instance | provisions KVM vms via libvirt | | maas | instance | bare-metal provisioning using Ubuntu MAAS | | packet | instance | provisions bare metal hosts on Packet | | rackhd | instance | bare-metal server provisioning via RackHD | | terraform | instance | creates resources using Terraform | | vagrant | instance | creates Vagrant VMs | | vsphere | instance | creates VMWare VMs |

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