Build actionable logs and alerts

Start with the incident questions

• Are users seeing errors?
• Which deployment is producing them?
• Which route and region are affected?
• Is an upstream dependency contributing to the latency?

1. Stream logs

apex logs tail --project checkout --environment production

Use the CLI for an interactive incident view or the API when you want to ship structured events into an observability platform.

2. Add meaningful fields

Include request ID, deployment ID, route, region, status code, and dependency timings. Avoid logging raw credentials, payment data, or unnecessary personal information.

3. Alert on impact

• Create a latency alert on p95 rather than average latency.
• Create an error alert scoped to the affected route or deployment.
• Use region-level thresholds when a global aggregate would hide a localized failure.

4. Close the loop

After an incident, save the alert and runbook next to the service documentation. The best alert reduces diagnosis time the next time the same failure appears.