edge-mc
Overview
edge-mc is a subproject of kcp focusing on concerns arising from edge multicluster use cases:
- Hierarchy, infrastructure & platform, roles & responsibilities, integration architecture, security issues
- Runtime in[ter]dependence: An edge location may need to operate independently of the center and other edge locations
- Non-namespaced objects: need general support
- Cardinality of destinations: A source object may propagate to many thousands of destinations.
Goals
- collaboratively design a component set similar to those found in the current kcp TMC implementation (dedicated Workspace type, scheduler, syncer-like mechanism, edge placement object definition, status collection strategy, and etc.)
- Specify a multi-phased proof-of-concept inclusive of component architecture, interfaces, and example workloads
- Validate phases of proof-of-concept with kcp, Kube SIG-Multicluster, and CNCF community members interested in Edge
Areas of exploration
- Desired placement expression: Need a way for one center object to express large number of desired copies
- Scheduling/syncing interface: Need something that scales to large number of destinations
- Rollout control: Client needs programmatic control of rollout, possibly including domain-specific logic
- Customization: Need a way for one pattern in the center to express how to customize for all the desired destinations
- Status from many destinations: Center clients may need a way to access status from individual edge copies
- Status summarization: Client needs a way to say how statuses from edge copies are processed/reduced along the way from edge to center.
Quickstart
TBD 🏗️
Next Steps
TBD 🏗️
Contributing
We ❤️ our contributors! If you're interested in helping us out, please head over to our Contributing guide.
Getting in touch
You may get in touch with us with the channels availabke in the kcp-dev community.