This is an incomplete work in progress, provided in the interest of sharing experience and gathering feedback. The code is pre-alpha quality right now. This means that it shouldn't be used in production at all.
The ClusterLink project simplifies the connection between application services that are located in different domains, networks, and cloud infrastructures.
For more details, see the document: ClusterLink extended abstract.
ClusterLink deploys a gateway into each location, facilitating the configuration and access to multi-cloud services.
The ClusterLink gateway contains the following components:
Control Plane
is responsible for maintaining the internal state of the gateway, for all the communications with the remote peer gateways by means of the ClusterLink CP Protocol (REST APIs), and for commanding the local DP to forward user traffic according to policies. Part of the control plane is the policy engine that can also apply network policies (ACL, load-balancing, etc.)Data Plane
responds to user connection requests, both local and remote, initiates policy resolution in the CP, and maintains the established connections. ClusterLink DP relies upon standard protocols and avoids redundant encapsulations, presenting itself as a K8s service inside the cluster and as a regular HTTP endpoint from outside the cluster, requiring only a single open port (HTTP/443) and leveraging HTTP endpoints for connection multiplexing.gwctl
is CLI implementation that uses REST APIs to send control messages to the ClusterLink Gateway.
The ClusterLink APIs use the following entities for configuring cross cluster communication:
- Peer. Represent remote ClusterLink gateways and contain the metadata necessary for creating protected connections to these remote peers.
- Exported service. Represent application services hosted in the local cluster and exposed to remote ClusterLink gateways as Imported Service entities in those peers.
- Imported service. Represent remote application services that the gateway makes available locally to clients inside its cluster.
- Policy. Represent communication rules that must be enforced for all cross-cluster communications at each ClusterLink gateway.
Here are the key steps for setting up your developer environment, making a change and testing it:
- Install Go version 1.20 or higher.
- Clone our repository with
git clone git@github.com:clusterlink-net/clusterlink.git
. - Run
make test-prereqs
and manually install any missing required development tools. - Run
make build
to ensure the code builds fine. This will pull in all needed dependencies. - If you are planning on contributing back to the project, please see our contribution guide.
ClusterLink can be set up and run on different environments: local environment (Kind), Bare-metal environment, or cloud environment. For more details, refer to the Installation Guide for ClusterLink.
ClusterLink can run in any K8s environment, such as Kind. To run the ClusterLink in a Kind environment, follow one of the examples:
- Performance example - Run iPerf3 test between iPerf3 client and server using ClusterLink components. This example is used for performance measuring. Instructions can be found Here.
- Application example - Run the BookInfo application in different clusters using ClusterLink components. This example demonstrates communication distributed applications (in different clusters) with different policies.Instructions can be found Here.
TBD
TBD
Our project welcomes contributions from any member of our community. To get started contributing, please see our Contributor Guide.
ClusterLink is intended to connect services and applications running in different clusters. As such, the project will implement or has implemented:
- Remote Service sharing
- Extending private Cloud service endpoints to remote sites
- Centralized management (future)
ClusterLink will be used in a cloud native environment with other tools. The following specific functionality will therefore not be incorporated:
- Certificate management: ClusterLink uses certificates and trust bundles provided to it. It does not manage certificate lifetimes, rotation, etc. - these are delegated to external tools.
- Enabling IP level connectivity between sites. ClusterLink uses existing network paths.
- Pod to Pod communications. ClusterLink works at the level of
Service
s (but you could create a Service per Pod Pod to Pod communications. ClusterLink works at the level ofService
s , but can support this scenario by creating a service per pod.
This project is licensed under Apache License, v2.0. Code contributions require Developer Certificate of Originality.
We follow the CNCF Code of Conduct.