OrdersApiLatencySLOBurn

# OrdersApiLatencySLOBurn

Larkspur Grocers is fictional. This note explains the alert. It is not an incident report. It does not record a page, a customer impact, or an MTTR.

## Signal

- Severity: `warning`
- Team: `orders`
- Pending duration (`for`): 2m

```promql
(
  slo:sli_latency_error:ratio_rate1h{service="orders-api",slo="latency"} > (14.4 * 0.01)
  and
  slo:sli_latency_error:ratio_rate5m{service="orders-api",slo="latency"} > (14.4 * 0.01)
)
or
(
  slo:sli_latency_error:ratio_rate6h{service="orders-api",slo="latency"} > (6 * 0.01)
  and
  slo:sli_latency_error:ratio_rate30m{service="orders-api",slo="latency"} > (6 * 0.01)
)
```

## What the condition means

The latency SLI is the share of orders-api requests slower than 300ms. The objective is 99 percent (budget 0.01). This alert uses the fast pair: 14.4x on 5m and 1h, or 6x on 30m and 6h.

## What an on-call would check

1. Read the expression above and the recording rules it uses.
2. Compare it with the promtool unit tests in `tests/rules/`. Every alert except Watchdog has a firing case and a non-firing case.
3. The annotation `runbook_url` is `https://runbooks.larkspur.example/OrdersApiLatencySLOBurn`. That host is a placeholder. This file is the copy in the repo.
4. This stack does not page anyone. Receivers use `*_file` credential paths and `example.invalid`.

返回告警目录