Master the art of cross-signal correlation in Grafana — link metrics to traces via exemplars, navigate from logs to traces via trace_id, and debug incidents using all three observability pillars together.
Understand why correlated signals are exponentially more powerful than isolated metrics, logs, and traces. Correlation turns three separate tools into one unified debugging experience.
cat <<'EOF'
=== WHY CORRELATION MATTERS ===
Without correlation (three isolated tools):
Metrics: "Error rate is 5% on order-api"
Logs: "Found 847 error logs in the last hour"
Traces: "There are slow traces for order-api"
Problem: You manually cross-reference timestamps and service names
With correlation (unified investigation):
1. Metrics dashboard shows error rate spike at 14:32
2. Click a data point → see example traces (exemplars)
3. Open a trace → see it hit payment-api with 5s latency
4. Click the payment-api span → see related logs
5. Logs show: "Stripe rate limit exceeded, retrying"
6. Root cause found in 2 minutes, not 2 hours
=== THE THREE CORRELATION LINKS ===
METRICS ──exemplars──→ TRACES
↑ |
| | trace_id
| ↓
time range LOGS
| |
└──── time range ───────┘
EOFIndividual signals answer different questions: metrics tell you something is wrong (what), logs tell you what happened (why), and traces tell you where the problem is (where). Correlation connects these answers into a single investigation flow. The three key links are: exemplars (metrics → traces), trace_id (logs ↔ traces), and time ranges (metrics ↔ logs). When all three are configured, you can start from any signal and navigate to the others without context-switching.
You see the contrast between isolated and correlated signals, followed by a diagram showing the three correlation links between metrics, logs, and traces.