Deploy and Configure Jaeger Tracing

Provides expert guidance for setting up, configuring, and deploying Jaeger distributed tracing in production - covering core architecture, Kubernetes deployment manifests, storage backend configuration, sampling strategies, OpenTelemetry instrumentation, performance tuning, security, and troubleshooting.

Use this skill when standing up a new Jaeger deployment, migrating from all-in-one to a production collector/query/storage architecture, choosing and configuring a storage backend (Elasticsearch or Cassandra), tuning sampling strategies for high-volume services, instrumenting an application with OpenTelemetry, or debugging missing traces, high latency, or storage issues in an existing Jaeger setup. Not a fit for tracing systems other than Jaeger/OpenTelemetry or for application-level logging unrelated to distributed tracing.

Covers the four core Jaeger components (agent, collector, query service, storage backend) and two deployment patterns: all-in-one for development, and a production architecture with separate collector, query, and storage, using collector clustering for high availability.

Provides ready-to-use Kubernetes manifests: a collector Deployment (3 replicas, Elasticsearch-backed, resource requests/limits defined) with its Service, and an Agent DaemonSet with hostNetwork enabled and UDP ports for span collection. Storage backend configuration is given for both Elasticsearch (TLS, index prefix, shard/replica counts, credentials via secret) and Cassandra (multi-node servers, keyspace, local datacenter, consistency level).

Sampling strategy is defined via a JSON service_strategies document supporting per-service probabilistic sampling (e.g. 0.1 for high-volume services, 1.0 for critical services) plus an adaptive default strategy with per-operation overrides (e.g. reduced sampling for health-check calls). OpenTelemetry integration includes a Go application instrumentation example initializing a TracerProvider with a Jaeger exporter and batcher.

Performance guidance covers collector batch sizing (1000-5000 spans), memory ballast, queue buffering, and concrete resource allocation recommendations per component (collector: 2-4 CPU / 4-8GB RAM; query service: 1-2 CPU / 2-4GB RAM; agent: 50m CPU / 128Mi RAM). Security guidance includes TLS flags for the collector gRPC endpoint and network policy recommendations. Monitoring is covered via a Prometheus ServiceMonitor example and a list of essential metrics (span ingestion rate/errors, storage health/latency, query response times, collector resource utilization). Troubleshooting includes common issues (missing traces, high latency, memory issues, storage problems) with debug commands for checking collector health, agent connectivity, and trace ingestion.

Deploys on Kubernetes (Deployment, DaemonSet, ServiceMonitor manifests) and Docker, integrates with Elasticsearch or Cassandra as storage backends, Prometheus for metrics collection, and OpenTelemetry (with a Go SDK example) for application instrumentation.

Platform and observability engineers deploying or operating Jaeger in production who need concrete Kubernetes manifests, storage backend configuration, sampling strategy design, and troubleshooting steps rather than a conceptual overview of distributed tracing.

kubectl exec -it jaeger-collector-xxx -- wget -qO- http://localhost:14269/