Documentationbreadcrumb arrow Grafana Cloudbreadcrumb arrow Telemetry signalsbreadcrumb arrow Use signals togetherbreadcrumb arrow Why correlation matters
Grafana Cloud

Why correlation matters

Each telemetry signal tells part of the story. When you correlate signals, you get the complete picture.

How correlation speeds troubleshooting

Without correlation, troubleshooting is slow:

  1. Alert fires (metrics).
  2. Manually search logs for related errors.
  3. Guess which trace might be relevant.
  4. Hope profiles are from the right time period.

With correlation, troubleshooting is fast:

  1. Alert fires (metrics).
  2. Click to see related logs.
  3. Jump to the exact trace.
  4. View the code-level profile.

Correlation turns hours of investigation into minutes.

How signals complement each other

Here’s how signals connect in a typical investigation:

How signals work together during an investigation

SignalWhat it tells youWhat it doesn’t tell you
MetricsSomething is wrong (error rate up)What went wrong
LogsWhat happened (error message)Where in the request flow
TracesWhere time was spent (slow service)Why that code is slow
ProfilesWhy code is slow (hot function)Business context

By moving between signals, you answer the complete question:

  1. Metrics: “Error rate increased at 2:15 PM.”
  2. Logs: “Database connection timeout errors.”
  3. Traces: “Database calls taking 30 seconds instead of 10ms.”
  4. Profiles: “Connection pool exhausted, threads waiting on locks.”

Common correlation patterns

Here are some common patterns you might follow when you are troubleshooting using metrics, logs, traces, and profiles.

Metrics to logs

Scenario: Alert fires for high error rate.

Correlation:

  1. Note the time and affected service from the alert.
  2. Query logs for that service and time range.
  3. Look for error messages or stack traces.

What connects them: Shared labels and attributes (service name, environment, instance).

Logs to traces

Scenario: Log shows an error for a specific request.

Correlation:

  1. Extract the trace ID from the log entry.
  2. Query Tempo for that trace ID.
  3. View the full request flow.

What connects them: Trace ID embedded in log entries.

Traces to profiles

Scenario: Trace shows a slow span in a service.

Correlation:

  1. Note the service and time range from the trace.
  2. View profiles for that service and time.
  3. Refer to what code was executing during that span.

What connects them: Service name and timestamp (or span profiles for direct linking).

Metrics to traces (exemplars)

Scenario: Latency metric shows a spike.

Correlation:

  1. View the latency metric with exemplars enabled.
  2. Click an exemplar point during the spike.
  3. Jump directly to the trace that caused it.

What connects them: Exemplars (trace IDs attached to metric samples).

Enable correlation in Grafana Cloud

To enable the correlation features described above, configure:

  • Shared labels and attributes across all signals (metrics and logs use labels; traces use attributes).
  • Trace context propagation in your application.
  • Derived fields for log-to-trace linking.
  • Exemplars for metrics-to-trace linking.
  • Traces to profiles for span-level linking.

For configuration steps, refer to Configure signal correlation or the individual product documentation:

Next steps