Signals: Traces, Metrics and Logs

Three signals, connected

Traces record the path of a request through services as a tree of spans with timings. Metrics are aggregated numbers over time (request counts, latency histograms, memory). Logs are timestamped records; OpenTelemetry can attach trace and span IDs to them so you can jump from a log line to its trace. All signals share a resource describing where they came from (service name, version, host, Kubernetes pod). The specification also covers profiling as a newer signal.

How the signals connect during an investigation

A typical workflow.

metrics   : error ratio for checkout jumps to 4%      -> alert fires
  |  exemplar / time range
traces    : slow, failing checkout traces show a 3 s span in payment-service -> card-gateway
  |  trace_id
logs      : payment-service log lines with that trace_id show "gateway timeout after 3000 ms"
resource  : all three carry service.name=payment-service, service.version=2.8.1, k8s.pod.name=...

Correlation is the main payoff

Shared IDs and resource attributes turn three separate tools into one investigation path.

त्वरित जाँच: Which signal shows how long each step of a single request took across services?

• Traces
• Metrics
• Resource attributes
• Configuration files