Topology by availability zone
catastrophic
severe
elevated
healthy
— select anything for evidence
Request path & failure propagation
internet → 4 × ECS edge task (public IPs, ECS-managed, 4/4 healthy)
↓ resolves app.acme.use2.internal — AAAA, TTL 15, multivalue, no health check
↓ 6 answers returned · 3 live · 3 point at terminated instances → ~50% connection failure
app tier (app-1/2/3, one per AZ)
↓ resolves auth.acme.use2.internal
↓ 4 answers returned · 0 live → auth unreachable by name, 100% failure
↓ resolves db-primary.acme.use2.internal
↓ 4 answers returned · 1 live → ~75% of writes fail; no failover to replica
↓ resolves cache → 1 answer, live, single instance in us-east-2a, no replica
egress: no NAT gateway. IPv4 outbound = S3 gateway endpoint only.
IPv6 outbound via eigw-0ab045701000f0e21. No SSM/interface endpoints → no Session Manager break-glass.
↓ resolves app.acme.use2.internal — AAAA, TTL 15, multivalue, no health check
↓ 6 answers returned · 3 live · 3 point at terminated instances → ~50% connection failure
app tier (app-1/2/3, one per AZ)
↓ resolves auth.acme.use2.internal
↓ 4 answers returned · 0 live → auth unreachable by name, 100% failure
↓ resolves db-primary.acme.use2.internal
↓ 4 answers returned · 1 live → ~75% of writes fail; no failover to replica
↓ resolves cache → 1 answer, live, single instance in us-east-2a, no replica
egress: no NAT gateway. IPv4 outbound = S3 gateway endpoint only.
IPv6 outbound via eigw-0ab045701000f0e21. No SSM/interface endpoints → no Session Manager break-glass.
Failure points ranked by blast radius
| # | Failure point | Radius if it fails | Recoverable? |
|---|