Application OpenTelemetry telemetry¶

Application traces, logs, and metrics emitted by your container are exposed through DataRobot's OpenTelemetry (OTel) observability surface. This surface is separate from the Workload API—the API itself owns Workload-scoped request statistics, lifecycle events, and per-replica status, and does not proxy OTel data. Instead, the platform aggregates whatever your container exports through the standard OTLP endpoint and renders it in the Console.

Container stdout and stderr are captured automatically and appear on Activity log > Logs at every lifecycle stage (including startup and crash diagnostics) without any SDK setup. The platform sets OTEL_EXPORTER_OTLP_ENDPOINT (and any auth env vars) on your container at start time, so when your application emits OTel signals via the SDK, they reach the same observability surface. Use explicit instrumentation when you want traces, application metrics, or structured, attribute-rich logs on top of automatic capture—for copy-ready Python SDK snippets, see Instrument a Workload with OpenTelemetry.

Where each signal surfaces¶

Each OTel signal type appears on a different Console tab:

Retention¶

OTel retention matches the Workload's artifact lifecycle status: 24 hours for draft artifacts and 30 days for locked artifacts. After retention elapses, OTel data is automatically deleted. See Monitoring concepts: Retention summary for the full retention model.