Module 17 cheat sheet
One-page incident-triage and interview revision summary for Daily-Work Playbooks and Career: first questions, evidence sources, Azure mapping, CLI commands, limits, decision rules, exam traps and cost traps.
Assumes: AWS CLI v2 · AWS SDK for .NET v4 · .NET 10 LTS · Region us-east-1 · facts as stated in the Module 17 lessons; items marked VERIFY there are not repeated as certain here. Verified against official sources on 2026-11-18.
A printable summary of Module 17. Each row links back to the playbook with the detail. For an incident, follow your bank's own process first; this page is a memory aid, not a procedure.
The method (every playbook)
Symptom, then first five minutes (scope, impact, contain), then evidence (logs, metrics, traces, CloudTrail), then ranked hypotheses, test one at a time, fix and verify, prevent recurrence, record for audit. Contain before you investigate. Never delete evidence during an incident.
Symptom to first look
| Symptom | First look | Playbook |
|---|---|---|
| Customer charged twice | Idempotency key and conditional write; queue consumer dedup | Duplicate payment |
| Batch misses its window | Per-stage timings, concurrency, data skew | EOD batch overrun |
| Calls to a partner time out | Egress path (NAT), DNS, TLS, timeouts vs retries | Partner API timeouts |
AccessDenied | Real caller, then the reason phrase, then the denying layer | IAM evaluation |
KMS AccessDeniedException | Key policy first, then IAM, grants, encryption context | KMS denied |
| ECS tasks restart | stoppedReason, exit code, task lifetime | ECS restarting |
| Lambda timeout or slow first call | Duration vs timeout, Init Duration, downstream latency | Lambda timeouts |
| 502 / 503 / 504 | Which hop returned it, target health, timeouts | 502/503/504 |
| Browser "CORS error" | Replay the preflight, read the real status | CORS |
| Bucket public | Block Public Access, then CloudTrail and access logs | Public bucket |
| Role over-privileged | Allowed vs last-accessed; staged trim | Role too broad |
| SQL slow after migration | Plans, statistics, parameter groups, pools, N+1 | Slow SQL |
| Messages stuck or repeated | Visibility timeout vs work time, receive count, DLQ | SQS |
| Rotation causes errors | Overlap period, cached secrets, refresh on failure | Key rotation |
| Who read this record? | Authority, subject, window; CloudTrail plus app audit log | Record access |
| Bill jumped | Account, service, usage type, resource, cause | Bill spike |
.NET / Azure to AWS
| Microsoft / Azure | AWS |
|---|---|
| Application Insights failures | CloudWatch Logs Insights, X-Ray / Application Signals |
| Activity Log | CloudTrail |
| RBAC access denied | IAM evaluation: identity, resource, boundary, SCP/RCP, session policy; explicit deny wins |
| Key Vault access policies | KMS key policy + grants + IAM |
| App Service diagnose and solve | ECS events and stopped tasks, ALB target health |
| Cost Management alerts | Budgets, Cost Anomaly Detection, Cost Explorer |
| Service Bus dead-letter | SQS DLQ + redrive |
Most useful CLI commands
| Task | Command |
|---|---|
| Who am I really? | aws sts get-caller-identity --profile learn-sandbox |
| Decode a denial message | aws sts decode-authorization-message --encoded-message TEXT |
| Simulate a policy | aws iam simulate-principal-policy ... (a hint, not proof) |
| Find a denied call | aws cloudtrail lookup-events --lookup-attributes ... |
| Run a Logs Insights query | aws logs start-query ... |
| ECS stopped task detail | aws ecs describe-tasks --cluster C --tasks ARN |
| Target health | aws elbv2 describe-target-health --target-group-arn ARN |
| Queue state | aws sqs get-queue-attributes --queue-url URL --attribute-names All |
| Redrive a DLQ | aws sqs start-message-move-task ... · aws sqs list-message-move-tasks ... |
| Contain a public bucket | aws s3api put-public-access-block ... |
| KMS rotation status | aws kms get-key-rotation-status --key-id ID |
| Force a secret rotation | aws secretsmanager rotate-secret --secret-id ID |
| Costs by service | aws ce get-cost-and-usage ... |
| Leftovers after a lab | aws resourcegroupstaggingapi get-resources --tag-filters Key=Project,Values=cea-lab |
Key limits and defaults
| Item | Value | Lesson |
|---|---|---|
| ALB idle timeout | 60 s default, configurable | 502/503/504 |
| ALB connect timeout to a target | 10 s | 502/503/504 |
| ALB Lambda target body limit | 1 MB (larger response gives 502) | 502/503/504 |
Kestrel KeepAliveTimeout | 130 s default, so longer than the ALB's | 502/503/504 |
| API Gateway REST integration timeout | 29 s default | 502/503/504 |
| Lambda timeout | 3 s default, 15 min maximum | Lambda timeouts |
| SQS message retention | 4 days default, up to 14 | SQS |
| SQS FIFO deduplication window | 5 minutes | SQS |
| CloudTrail event history | Management events, 90 days | IAM evaluation |
| CloudTrail data events | Off by default; you can only show what was logged | Record access |
| KMS key deletion wait | 7 to 30 days, irreversible after | Key rotation |
| Block Public Access | Four settings, per bucket and per account | Public bucket |
Decision rules
| Situation | Rule |
|---|---|
| Unsure it is safe to retry | Only if the server can recognise the retry (key per intent, atomic claim, replay) |
| Reversing a payment | Approved, idempotent, linked correcting entry; never edit |
| Visibility timeout vs work time | Timeout must exceed the longest handler run, or messages reappear mid-work |
| ECS task dies in seconds | Startup crash: read the first log lines and the secrets |
| ECS task lives about grace plus a minute | Health check failing: fix port, path, status or security group |
| Exit code 137 | SIGKILL: memory over limit, or slow shutdown past stopTimeout |
| Cross-account denial | Both the caller's policy and the resource or key policy must allow |
| Trimming a role | Attach the new policy before removing the old; watch a full business cycle |
| Rotating a database secret | Alternating users, with clients that refresh on login failure |
| Public bucket | Contain with Block Public Access first; engineers give facts, the DPO decides on notification |
| No logs for a window | The answer is "unknown", not "nobody" |
Common exam and interview traps
| Trap | Reality |
|---|---|
| "The simulator said allowed" | It misses real-world factors; confirm with the real error and CloudTrail |
| "IAM allow is enough for KMS" | The key policy must permit the account or the principal first |
| CORS "fixed" with a wildcard | Not with credentials; use an exact origin list, one layer owning CORS |
| 401 shown as CORS error | Error responses need CORS headers too; authorizers must not guard OPTIONS |
| Raise Lambda timeout past the API limit | The API still returns 504; fix the slow path |
| Raising memory for an ECS loop | Classify the stop first; memory is only one of six classes |
| CloudTrail names the person | It names the role; the application audit log must name the person |
| Systems Manager Incident Manager or CloudTrail Lake | Closed to new customers; do not design on them |
Cost-trap list
| Trap | Where to look |
|---|---|
| NAT gateway data processing | Usage type by bytes; add gateway or interface endpoints |
| Idle load balancers and forgotten instances | A step up that stays flat in the daily graph |
| Log ingestion and groups with no retention | CloudWatch usage types; set retention |
| Retry or event loops | A ramp or spike that tracks request volume |
| Versioning without lifecycle, old snapshots | Storage byte-hours growing |
| Anything left after a lab | Tag-based sweep above; teardown within the hour |
Independent learning resource — not affiliated with, endorsed or sponsored by Amazon Web Services. AWS and related marks are trademarks of Amazon.com, Inc. or its affiliates. All banking data in examples is fictional. Sample architectures are for learning and are not automatically compliant with GDPR, PCI DSS or any banking regulation.