Parent: backend#830 (RFC-0001 epic) · RFC: §8.5 · Flow family: backend#877 (Connect / install) · Phase 1 — CLI
Related: cli#5 (split stdout/stderr for --output json — adjacent), the installer reorder backend#838 (streams the rollout)
Why
§8.5: the connect flow is zero-prompt on a headless box, so the failure path has to stand on its own — otherwise "easy when it works" becomes "stuck when it doesn't." Today the CLI has no verbosity flag (only --plain), and cluster doctor checks cluster health only, so it can't diagnose a failed provision.
Scope
--verbose / TRACEBLOC_LOG_LEVEL that streams the device-flow → provision → Helm steps (default stays quiet — ~6 ✔ lines).- Always write a persistent
~/.tracebloc/install-<ts>.log.
- On any failure: print the exact resume command (re-running is idempotent — §7.2/§8.2) and point at
tracebloc cluster doctor.
- Extend
cluster doctor to also check auth / config / token state (today it's cluster-health only).
Acceptance
--verbose streams the staged steps; default output stays quiet.- Every run writes an
install-<ts>.log.
- A simulated failure prints a copy-pasteable resume command + the doctor hint.
cluster doctor reports auth/config/token problems, not just cluster health.
Notes
Installer-side §8.5 work (streamed per-Deployment rollout + "still pulling images" heartbeat; 🟢 Connected = a real readiness probe, not a sleep) lives with the installer reorder (backend#838) — cross-ref, not this CLI ticket's scope.
Work Type: Feature · Squad: DevEx · Area: SDK (CLI) · Priority: P1
Parent: backend#830 (RFC-0001 epic) · RFC: §8.5 · Flow family: backend#877 (Connect / install) · Phase 1 — CLI
Related: cli#5 (split stdout/stderr for
--output json— adjacent), the installer reorder backend#838 (streams the rollout)Why
§8.5: the connect flow is zero-prompt on a headless box, so the failure path has to stand on its own — otherwise "easy when it works" becomes "stuck when it doesn't." Today the CLI has no verbosity flag (only
--plain), andcluster doctorchecks cluster health only, so it can't diagnose a failed provision.Scope
--verbose/TRACEBLOC_LOG_LEVELthat streams the device-flow → provision → Helm steps (default stays quiet — ~6 ✔ lines).~/.tracebloc/install-<ts>.log.tracebloc cluster doctor.cluster doctorto also check auth / config / token state (today it's cluster-health only).Acceptance
--verbosestreams the staged steps; default output stays quiet.install-<ts>.log.cluster doctorreports auth/config/token problems, not just cluster health.Notes
Installer-side §8.5 work (streamed per-Deployment rollout + "still pulling images" heartbeat;
🟢 Connected= a real readiness probe, not a sleep) lives with the installer reorder (backend#838) — cross-ref, not this CLI ticket's scope.Work Type: Feature · Squad: DevEx · Area: SDK (CLI) · Priority: P1