Back to Cookbook
KiloClaw

Distributed Trace Hunt

Find missing spans and broken propagation in microservice flows

Debug microservices by enforcing end-to-end correlation (trace IDs, request IDs) and systematically locating propagation breaks or sampling gaps.

CommunitySubmitted by CommunityWork20 min

INGREDIENTS

🐙GitHub🔎Web Search

PROMPT

Create a skill called "Distributed Trace Hunt". Ask me for: - The request journey (start → end) and involved services - Current tracing stack (OpenTelemetry/Jaeger/etc.) and sampling approach - One example trace/span screenshot or IDs (if available) Output: - A step-by-step trace propagation audit checklist - Likely root-cause buckets and how to confirm each - A minimal staging test plan that asserts end-to-end spans exist

How It Works

In distributed systems, partial traces are common. This recipe builds a repeatable

trace-driven debugging path: define the journey → validate propagation → close gaps.

Triggers

  • Traces contain only some services (missing spans)
  • You can't follow a single request end-to-end
  • Production incidents require correlating logs/metrics/traces quickly

Steps

  1. Choose one representative request journey and record its expected service chain.
  2. Verify propagation headers and context injection/extraction per hop.
  3. Check sampling policy and exporter/backpressure settings.
  4. Add a "trace assertion" in a staging smoke test: request should produce spans in all hops.
  5. Create a playbook for "missing spans" triage: sampling vs instrumentation vs async boundaries.

Expected Outcome

  • You can trace a request through the intended service chain reliably.
  • Incidents become faster to debug and less dependent on tribal knowledge.

Example Inputs

  • "Our traces show the API gateway and service A, but nothing after."
  • "Async job continuation loses the trace context."
  • "We need to correlate frontend RUM with backend traces."

Tips

  • Treat broken propagation as a correctness bug, not a logging preference.
Tags:#debugging#observability#microservices#performance