jaeger

Best Practices

• Configure per-service sampling strategies — avoid using one rate for all services
• Use probabilistic sampling for high-throughput services and rate-limiting for low-traffic ones
• Set sampling rate to 0.0 for health-check and readiness probe operations to reduce noise
• Use Elasticsearch or Cassandra for production storage backends — never use in-memory for production
• Enable OTLP receiver (gRPC :4317, HTTP :4318) for OpenTelemetry SDK compatibility
• Configure batch processor with appropriate timeout and send_batch_size for collector throughput
• Set memory_limiter processor to prevent OOM in high-throughput environments
• Use TLS for collector endpoints in production (grpc.tls, http.tls)
• Set span storage TTL (max_span_age) for automatic data expiration and storage management
• Configure index_prefix for Elasticsearch to avoid index name conflicts with other applications