// the find
karmada-io/karmada
Open, Multi-Cloud, Multi-Cluster Kubernetes Orchestration
Karmada is a CNCF-graduated control plane for running one set of Kubernetes manifests across multiple clusters and clouds. It's aimed at platform teams who already run several clusters (multi-region, multi-cloud, or hybrid) and want a single API to schedule, fail over, and rebalance workloads between them instead of hand-rolling that logic per app.
It reuses native Kubernetes resource definitions and just adds new CRDs (PropagationPolicy, OverridePolicy) on top, so existing manifests and tooling don't need to change. The controller split (Cluster, Policy, Binding, Execution) gives a traceable path from 'policy applied' to 'resource landed in member cluster', which makes debugging propagation failures tractable instead of a black box. CNCF graduation plus visible OpenSSF Scorecard, CII Best Practices, and Codecov badges signal actual governance and security review, not just a checkbox project. The README's compatibility matrix testing against 10+ Kubernetes versions is the kind of unglamorous maintenance work a lot of infra projects skip.
You're standing up an entire second control plane (its own etcd, API server, controller manager, scheduler) just to coordinate other clusters, which is meaningful operational surface for a team running only 2-3 clusters. The two-kubeconfig-context model (karmada-apiserver vs karmada-host) and the new PropagationPolicy/OverridePolicy vocabulary mean the 'zero change' pitch applies to your workloads, not to the people operating the platform, who have real ramp-up ahead. The README says nothing about RBAC or tenancy boundaries around who can write propagation policies, which matters a lot once more than one team shares a Karmada instance. Community meetings split across Chinese and English tracks hint at a contributor base concentrated in specific companies/regions, worth a quick look at MAINTAINERS.md before treating this as vendor-neutral infrastructure.