What problem does it solve? Running day-2 operations on self-hosted RKE2 clusters requires knowing RKE2-specific paths, socket locations, and controller behaviors that differ from stock Kubernetes, and getting them wrong causes TLS failures, empty container listings, or broken GPU workloads. ## Core Features & Use Cases - Cluster Access and Tooling: Locate the admin kubeconfig at /etc/rancher/rke2/rke2.yaml, configure tls-san entries for external access, and use bundled kubectl, crictl, and ctr against the k3s-style containerd socket. - Debugging and Metrics: Find logs for rke2-server, rke2-agent, containerd, kubelet, and static control-plane pods, and enable supervisor metrics on port 9345 including certificate expiration and loadbalancer health gauges. - Add-on and GPU Management: Deploy applications via the manifests directory, HelmChart and HelmChartConfig CRDs, and run the NVIDIA GPU operator with the correct containerd socket configuration across operator versions. - Use Case: A platform engineer joins a new agent node, notices pods stuck pending, uses crictl with the correct socket to inspect containers, checks kubelet logs, and resolves a Canal IP exhaustion issue by clearing stale lock files. ## Quick Start Ask the assistant to show how to access an RKE2 cluster with kubectl and inspect running containers using the bundled crictl binary.
