What are Prometheus Recording Rules?

Question

Accepted Answer

**Recording Rules** allow **pre-computing frequently used metrics** and saving them as new time series.

**Simple analogy:** Recording Rules are like web caching: instead of computing a complex query each time, you compute it in advance and save the result.

**Why:**
- Faster queries (no real-time computations).
- Reduced load on Prometheus.
- Simplified alerts (simple metric instead of complex PromQL).

**Example rules file:**
```yaml
groups:
  - name: http_rules
    interval: 30s
    rules:
      - record: http_requests:rps5m
        expr: rate(http_requests_total[5m])

- record: http_errors:ratio5m
        expr: |
          sum(rate(http_requests_total{status=~"5.."}[5m])) /
          sum(rate(http_requests_total[5m]))

- record: http_latency:p95_5m
        expr: histogram_quantile(0.95, rate(http_request_duration_seconds_bucket[5m]))

- record: pod_cpu:usage_rate
        expr: rate(container_cpu_usage_seconds_total[5m])

- name: slo_rules
    rules:
      - record: service:availability5m
        expr: |
          sum(rate(http_requests_total{status=~"2.."}[5m])) /
          sum(rate(http_requests_total[5m]))
```

**Usage in alerts:**
```yaml
groups:
  - name: alerts
    rules:
      - alert: HighErrorRate
        expr: http_errors:ratio5m > 0.05
        for: 5m
        labels:
          severity: critical
```

**Naming convention:** `level:metric:operations`

DevOps: What are Prometheus Recording Rules?