Observability

NPipeline provides three complementary observability dimensions, each available as a separate extension package:

• Metrics - per-node work timing, input-wait timing, throughput, and optional memory metrics. For lazy stream nodes with per-node observability enabled, timing is finalized at dataflow completion. Pluggable sinks let you route metrics to logging, Prometheus, or any custom backend. Note: INodeMetrics.DurationMs is work duration; use WallDurationMs for wall-clock elapsed-time comparisons.
• Tracing - OpenTelemetry integration that creates Activity spans for each pipeline run and node execution, compatible with Jaeger, Zipkin, and Azure Monitor.
• Lineage - data provenance tracking with per-item correlation IDs, sampling, and overflow policies for production use.

All three are opt-in and add zero overhead when not configured.

In This Section

• Metrics and Monitoring - collect per-node and per-pipeline metrics with pluggable sinks
• OpenTelemetry Integration - distributed tracing with standard exporters
• Data Lineage - track data provenance with per-item correlation and sampling