ReadMeFeed / Community knowledge
Real questions. Useful conversations. Find the people who know your stack.
A decision guide that compares head‑based and tail‑based sampling strategies, outlines trade‑offs, and shows how to configure and validate each option in a high‑throughput system.
Learn how to use Jaeger’s probabilistic and rate‑limiting sampling to keep observability high while controlling storage, and see a concrete example that shows how to configure and verify the settings.
Jaeger offers constant, probabilistic, rate-limiting, and adaptive sampling. Here's how to choose between them, verify what clients actually received, and when you need tail sampling instead.
Jaeger Tracing Sampling A configuration utilizing a static local sampler functions correctly in a local development environment. Production Sampling Failure This same setup results in zero sampled traces when deployed to production infrastructure. What is the documented fallback behavior of Jaeger client libraries when the remote sampling endpoint is unreach
Goal Identify a latency bottleneck in a microservice architecture before applying costly optimizations. Constraints High‑throughput traffic forces the use of adaptive sampling to avoid collector overload. Clock skew between services can distort span timestamps in the Jaeger UI. Jaeger’s UI visualizes only the sampled traces, potentially hiding short‑duration
In Jaeger, the Collector ingests spans via gRPC or Thrift and forwards them to a storage backend. The Query service then retrieves these spans and the UI renders a timeline view. A documented feature is the sampling mechanism, where the Collector drops a configurable percentage of traces to limit storage usage. This introduces a tension: higher sampling rate
Remote sampling configuration scope I am configuring centralized remote sampling for Jaeger, where client libraries poll the collector's /sampling endpoint for their strategy. The strategy is defined in a file mounted into the collector (via a ConfigMap in a Kubernetes/Operator deployment), and the goal is for all services to adopt updated per-service probab
I'm designing the tracing pipeline for a set of services that will export OpenTelemetry spans through a Collector, and I need to pick a sampling strategy before rollout. The two documented options pull in opposite directions. With head-based sampling (for example a ParentBased sampler wrapping TraceIdRatioBased in the SDK), the decision happens at span creat
Problem OpenTelemetry auto‑instrumentation libraries default to a probability sampler (often 10% for Java, 5% for Python) to limit trace volume. The SDK exposes a sampler configuration API, but changing the sampler at runtime requires code changes or a restart; no cross‑language, environment‑variable override is defined in the core spec. Vendor exporters (Ja
opentelemetry
jaegertracing