Telemetry
LyftData exposes several signal families. Choose the family that answers your question instead of treating every signal as one telemetry stream.
Workload observability
Section titled “Workload observability”Workload observability stays with the LyftData deployment or a destination you configure. It helps operators understand the control plane, workers, jobs, and runtime:
| Signal | Use it for | Start here |
|---|---|---|
| Logs and issues | Historical events, warnings, errors, and current problems | Logs and Issues |
| Metrics | Rates, counts, resource trends, dashboards, and alerts | Monitoring |
| Distributed traces | Following bounded activity across a deployed job and worker | Distributed Traces |
| Runtime evidence | Inspecting component state, artifacts, and operation-specific receipts | Runtime Evidence and Receipts |
| Health and support diagnostics | Checking server, worker, storage, version, and support posture | Troubleshooting |
Run & Trace in the job editor is a transient authoring and validation tool. It is related to runtime tracing, but it does not prove sustained performance, distributed placement, or production recovery.
OpenTelemetry inputs and outputs are job components for receiving or exporting supported workload signals. Their delivery behavior follows the selected component and topology.
Product telemetry
Section titled “Product telemetry”Product telemetry is the separate reporting path that sends installation identity and bounded operational summaries to a configured telemetry receiver. It has its own authentication, TLS, storage, retention, access, deletion, and privacy controls.
It is not workload observability or OpenTelemetry export. Disabling product telemetry does not disable local logs, metrics, traces, health checks, or operator dashboards. Review the payload and controls in Product Telemetry.
Access and data handling
Section titled “Access and data handling”Observability endpoints use route-specific authentication and authorization. Use a dedicated read-only principal for dashboards and collectors. If you use the optional network allowlist for selected worker read endpoints, restrict it to exact collector addresses and review Access Control.
Logs, traces, runtime artifacts, and support bundles can contain sensitive workload or operational metadata. Apply least privilege, retention, redaction, and export controls to each signal family. A signal being visible in the UI does not mean it is safe to attach to a ticket.
When a signal is missing
Section titled “When a signal is missing”Source support for a signal does not prove that it is enabled or complete in a specific installation. Check:
- selected product edition and build features;
- server and worker versions;
- worker connectivity and exact target-job serviceability;
- collection, sampling, and retention settings;
- tenant and environment filters;
- exporter or destination health; and
- the time range used by the UI or query.
If live UI status appears stale, check long-lived HTTP connections through reverse proxies, VPNs, and corporate gateways. Then correlate the result with the worker and job status rather than relying on one signal alone.