Canonical archetype explorer
Read-only reference contracts plus the validator-trace preview. Pick a defended invariant to see the trace fire on a known violation.
Order events Lambda silently dropping logs
SEV-2 · highMedium≈ 30 minawsEventBridge invokes a Lambda. Metrics look healthy. Downstream queue receives events. CloudWatch Logs show nothing. Find the observability asymmetry and restore logs without compromising the security posture.
Customer support reports a confusing intermittent issue: order confirmation emails are sometimes delayed by up to an hour. The on-call engineer confirms the chain step by step. Events are landing in EventBridge. The `order-events-publisher` Lambda is invoked. The function's CloudWatch metrics show healthy invocation count and zero errors. SQS messages reach the downstream queue. By every observable system, the pipeline is healthy. But the support team is asking for logs. The Lambda's log group `/aws/lambda/order-events-publisher` exists in CloudWatch. It has no log streams. No log events have arrived since the function was first deployed. The team needs to know why the Lambda is succeeding *and* invisible — and restore observability without making the system less safe.
Business impact: Customer support cannot investigate intermittent order delays. Engineering cannot root-cause without logs. Each missed log line costs roughly 12 minutes of investigation time per incident, currently running at 8 incidents per week. At $180/hr engineer time that is approximately $3,600/week of analytical debt — and an SLA target for support-response-time at risk.