sarulon / gravity Goto Github PK

Gravity is an upstream Kubernetes packaging solution that takes the drama out of deploying and running applications in someone else's cloud accounts, on-premise data centers, edge locations and other "uncharted territory" environments.

With Gravity, your Kubernetes apps can run and be regularly updated anywhere in the world without a massive DevOps team.

Gravity is an open source toolkit for creating "images" of Kubernetes clusters and the applications running inside the clusters. The resulting images are called cluster images and they are just .tar files.

A cluster image can be used to re-create full replicas of the original cluster in any environment where compliance and consistency matters, i.e. in locked-down AWS/GCE/Azure environments or even in air-gapped server rooms. An image can run without human supervision, as a "kubernetes appliance".

Gravity has been running in production in major financial institutions, government data centers and enterprises. Gravitational open sourced it in the fall of 2018.

There are plenty of Kubernetes distributions out there. Most of them aim to be flexible, general purpose platforms. Gravity has a more narrow focus on compliance and reducing the overhead of managing Kubernetes:

• Gravity clusters are idempotent, i.e. clusters created from the same image are always identical. There is no configuration drift over time; no "special snowflakes".
• Gravity clusters are always "wrapped" with a privileged access gateway called Teleport, which unifies k8s and SSH authentication, integrates with SSO and keeps a detailed audit log for compliance purposes. It even records the interactive SSH and kubectl exec sessions.
• Gravity clusters deployed world-wide can be remotely managed via built-in reverse SSH tunnels, i.e. developers can have access to thousands of k8s API endpoints even if they're located behind NAT/firewalls.
• Gravity includes tools to perform infrastructure validation prior to cluster provisioning. This allows cluster designers to prevent users from installing clusters on infrastructure that does not meet the system requirements.
• Gravity clusters only allow Kubernetes components that have been thoroughly tested by Gravitational Inc for compatibility and stability. These components are called a "Kubernetes Runtime". Users can pick a Runtime but Gravity does not allow any customization of individual components of Kubernetes.

We have seen the following primary use cases for using a image-based Kubernetes approach (there may be others):

• Deploying and running complex SaaS applications into on-premises enterprise environments.
• Deploying and running complex SaaS applications in thousands of edge locations (retail, transportation, energy, etc).

Anyone who needs Kubernetes best practices out of the box, without having to proactively manage it can benefit from Gravity. It allows you to focus on building your product instead of managing Kubernetes.

A Cluster Image produced by Gravity includes:

• All Kubernetes binaries and their dependencies.
• Built-in container registry.
• De-duplicated layers of all application containers inside a cluster.
• Built-in cluster orchestrator which guarantees HA operation, in-place upgrades and auto-scaling.
• Installation wizard for both CLI and web browser GUI.

An image is all one needs to re-create the complete replica of the original Kubernetes cluster, with all deployed applications inside, even in an air-gapped server room.

Take a look at the examples directory in this repository to find examples of how to package and deploy Kubernetes applications using Gravity.

The following examples are currently available:

• Wordpress. Deploys Wordpress CMS with an OpenEBS-backed persistent storage.

A cluster image created with Gravity can be used for:

1. Creating many Kubernetes clusters from scratch, on any infrastructure.
2. Installing applications contained in the cluster image into an existing Kubernetes cluster, like OpenShift.

Developers can continuously update their applications using different methods:

1. Vanilla CI/CD using Kubernetes APIs, which is available for every cluster. This is probably what you're already doing.
2. Via "polling model", when each Gravity cluster will automatically download updates from a Gravity Hub, letting cluster users decide when/if they want to upgrade. This method is recommended for traditional on-premise environments when developers do not have access to each deployment site.
3. Offline method, when a developer prepares a new cluster image which can be distributed via offline media. This method is suitable for air-gapped environments.

Each cluster provisioned with Gravity includes the built-in SSH/Kubernetes gateway called Teleport. Teleport provides the following benefits:

• One-step authentication which issues credentials for both k8s API and SSH.
• Ability to implement compliance rules like "developers must never touch production data".
• Ability to grant remote access to the cluster via SSH or via k8s API, even if the cluster is located behind NAT with no open ports.
• Keeps a detailed audit log (including fully recorded interactive sessions) for all SSH commands and all kubectl commands executed on cluster nodes.

Teleport can also be used independently without Gravity, it has been audited multiple times by reputable cyber security companies and it has been deployed in production in multiple organizations.

Yes! Even though Gravity was open sourced in September 2018, it started life much earlier, as a component of a larger, proprietary system called Telekube.

Fully autonomous Gravity clusters are running inside of large banks, government institutions, enterprises, etc. Some of the commercial users of Gravity are listed on the Gravitational web site

Gravity was built by Gravitational Inc, a company based in Oakland, California.

The original use case for Gravity was to allow Kubernetes applications to be deployed into 3rd party environments, like on-premises datacenters. That's why Gravity includes features like the built-in, graphical cluster installer, infrastructure validation and a built-in privileged access manager (Teleport) for providing remote support.

These features also resonated with security-minded teams who need to run applications in environments where compliance matters. Gravity clusters are always identical and do not allow any configuration drift over time. This allows cluster architects (aka, Devops or SREs) to "publish" clusters that are approved for production and allow multiple teams within the organization to rapidly scale their Kubernetes adoption without having to become security and Kubernetes experts themselves.

Gravity is written in Go. There are two ways to build the Gravity tools from source: by using locally installed build tools or via Docker. In both cases you will need a Linux machine.

Building on MacOS, even with Docker, is possible but not currently supported

$ git clone [email protected]:gravitational/gravity.git
$ cd gravity

# Running 'make' with the default target uses Docker.
# The output will be stored in build/current/
$ make

# If you have Go 1.10+ installed, you can build without Docker which is faster.
# The output will be stored in $GOPATH/bin/
$ make install

# To remove the build artifacts:
$ make clean

While the code is open source, we're still working on updating the documentation to reflect the differences between the proprietary and community/OSS editions of the software. We are also working on providing open source users with pre-built binaries on a regular basis.

To contribute, please read the contribution guidelines.

• Want to join our team to hack on Gravity? We are always hiring!
• Want to stop managing Kubernetes and have autonomous appliance-like clusters?
• Want to take your complex SaaS application and convert it into a downloadable appliance so your customers can run it on their own AWS account or in a colo?

Reach out to [email protected]