OpenTelemetry Tracing and Prometheus Metrics in Hive Router
Hive Router now includes first-class observability with OpenTelemetry: distributed tracing, configurable metrics, OTLP exporters, and a Prometheus scrape endpoint.
Tracing with OpenTelemetry
Hive Router can emit traces to Hive Console Tracing and any OTLP-compatible backend over HTTP or gRPC.
Tracing includes GraphQL-specific spans as well as router and subgraph lifecycle spans, with configurable propagation for trace_context, baggage, b3, and jaeger.
Metrics with OTLP and Prometheus
Hive Router emits OpenTelemetry metrics covering inbound and outbound HTTP traffic, GraphQL errors and execution, supergraph polling and processing, and internal cache activity.
Metrics can be exported to OTLP-compatible backends, exposed for Prometheus scraping, or both. For Prometheus, you can expose metrics on the router's HTTP server or a dedicated port, with a configurable path (default /metrics).
Highly customizable
Tracing and metrics can be dialed in to balance cost and throughput:
- Sampling and batching - supports head-based and parent-based sampling, with configurable batch processing and export settings
- Metric instrument overrides - disable selected instruments or drop high-cardinality labels
- Histogram aggregation controls - supports
explicitandexponentialaggregations, with dedicated buckets for time and size units
This gives you the flexibility to keep observability lightweight at the start, and scale it up as your traffic and requirements evolve.