Symptom: Errors mentioning no matches for kind, no kind is registered, or data.fluid.io/v1alpha1 not found.
Fix: Install Fluid on the cluster before using fluid. See the Fluid install guide.
kubectl get crd datasets.data.fluid.ioSymptom: Timeouts, connection refused, or TLS errors when running fluid inspect or fluid diagnose.
Fix: Verify kubeconfig and cluster health:
kubectl cluster-info
kubectl get nsEnsure --context and --kubeconfig point at the intended cluster.
Symptom: Errors about non-interactive mode when using default -o tui or omitting the dataset name on inspect.
Fix: Pass explicit arguments and a non-TUI output mode:
fluid inspect my-dataset -n default -o table
fluid diagnose my-dataset -n default -o dirSymptom: Forbidden or cannot get resource when listing or collecting resources.
Fix: Your kubeconfig user or ServiceAccount needs read access to Fluid CRDs, Pods, Events, PVCs, PVs, and related objects in the target namespace (and fluid-system if using --include-controller-logs).
Symptom: Message like completed with N partial collection issues.
Fix: Open summary.txt and manifest.json in the output directory. Failed entries include a reason (e.g. pod not found, log stream error). The bundle is still useful; re-run with --since or --no-logs on very large clusters if collection is slow.
Generated command reference lives in docs/reference/. After changing commands or flags in cmd/, run:
make docsand commit the updated files. CI verifies they match the Cobra tree.
- Command help:
fluid <command> --help - Inspect guide / Diagnose guide
- GitHub issues — include
fluid version, exact command, and redacted output