将控制编码化
先测量问题。编写按 error 过滤并按 service 分组的 Logs Insights 查询，使首要错误源在几秒内可见。 最常见的错误是未确认门禁能捕获你关心的问题就先加上。在接入 CI 之前，先对当前代码库运行 aws logs start-query –log-group-name /aws/lambda/api –query-string ‘fields @timestamp, @message | filter level == “error” | stats count() by service’ 并检查输出。
aws logs start-query --log-group-name /aws/lambda/api --query-string 'fields @timestamp, @message | filter level == "error" | stats count() by service'检测漂移
门禁就位后，让失败可操作。aws logs start-query –log-group-name /aws/lambda/api –query-string ‘fields @timestamp, @message | filter level == “error” | stats count() by service’ 失败时应指向具体文件和行号，而非泛泛的错误。在锁文件中固定工具版本，防止上游发布悄悄改变行为。每次拉取请求都应像审查测试结果一样审查门禁输出。
验证
在本地运行 aws logs start-query –log-group-name /aws/lambda/api –query-string ‘fields @timestamp, @message | filter level == “error” | stats count() by service’ 确认通过后推送。CI 中门禁应阻断不合规的拉取请求。若人为绕过门禁直接推送，应在下次定期扫描中被发现。
注意事项
- The gate only runs on events that trigger the workflow; locally pushed commits bypass it.
- Configuration drift is possible if someone edits settings outside version control.