Fleet is Rancher/SUSE’s GitOps engine for managing Kubernetes at scale — specifically, at the “thousands of clusters” scale that edge and retail deployments produce. It’s the continuous delivery backbone of Rancher 2.5+ and ships as the default app-delivery mechanism for RKE2, K3s, and Rancher-managed clusters.
Fleet has a two-level architecture. A central “upstream” Kubernetes cluster runs the Fleet controller and holds GitRepo and Bundle resources; downstream clusters run a lightweight agent that registers with the upstream and pulls its assigned bundles. A GitRepo points at a Git repo containing raw manifests, Helm charts, or Kustomize overlays; Fleet renders them into Bundles, then uses ClusterGroup label selectors to decide which downstream clusters get which bundle, with per-target customizations. Because the targeting is label-based, rolling a config change across 10,000 edge stores is the same operation as rolling it across one.
Compared to Argo CD and Flux, Fleet’s distinguishing trait is that scale-out model: one control plane managing many heterogeneous clusters, with the agent-pull design tolerating flaky edge networks. The tradeoff is a smaller ecosystem and a UX that’s tightly coupled to Rancher. If you already run Rancher, Fleet is what you use; if you don’t, Argo CD or Flux are usually the first picks.