OnCallReadyStart learning

AWS V: Data, Resilience & the Edge

Managed databases and what happens when they fail: RDS on real PostgreSQL (Multi-AZ, read replicas, failover and the connection pools that miss it, snapshots and point-in-time restore, parameter groups, RDS Proxy) and Aurora, DynamoDB keys, capacity and hot partitions, ElastiCache, EFS and S3 storage classes, replication and Object Lock, AWS Backup plans, vaults and restores against an RPO, Route 53 routing policies and health checks, CloudFront with OAC and WAF, and many VPCs joined by peering, Transit Gateway and VPN - against the same simulated account, with real outputs and real errors.

In plain words

Picture the shop again, but this time look at the parts that remember things and the parts that face the street. The filing room is the database: a clerk, a twin clerk in another building who copies every change, photographs of the files every night, and a safe that nobody can open early. Out front there are signposts that send customers to an open branch, kiosks near the customers that keep copies of the catalogue, a guard who reads every letter at the door, and corridors and private roads that join the buildings.

On AWS those are RDS, DynamoDB, ElastiCache, EFS and S3 with their copies and backups, then Route 53, CloudFront, WAF, Transit Gateway and VPN. This chapter shows how each one fails, and what keeps it alive.

Why it matters on call

State is where incidents hurt most. A web server that dies is replaced in minutes; a database that failed over while the connection pools kept dead connections, a backup nobody had restored, or a lifecycle rule that quietly deleted the logs you need, turns into hours and lost data.

The edge is the other half: the DNS answer, the cached page and the firewall rule sit in front of everything, so their mistakes look like the whole site is down. The chapter teaches each service through the failure you will meet on call, and the habit that ties it together: a copy only counts once a restore or a failover test has proved it.

Lessons

  1. 38.1RDS: a managed PostgreSQL you reach by DNS
  2. 38.3Multi-AZ, read replicas and the failover your clients must survive
  3. 38.8RDS backups: snapshots, point-in-time restore and what RPO really costs
  4. 38.10DynamoDB: design for the questions you will ask
  5. 38.13DynamoDB in production: capacity, throttling, hot keys, TTL and backups
  6. 38.15ElastiCache: when a cache helps, and what happens when it fails over
  7. 38.17Block, file and object: EBS, instance store, EFS and when each fits
  8. 38.19S3 storage classes, lifecycle and Glacier restores
  9. 38.22S3 replication and Object Lock: copies elsewhere, copies nobody can delete
  10. 38.24AWS Backup: plans, vaults, copies, locks, and restores you have actually tested
  11. 38.28Route 53 routing policies and health checks: DNS that answers per query
  12. 38.31CloudFront: caching at the edge, OAC, invalidations and TLS
  13. 38.35AWS WAF and Shield: filtering requests before they reach you
  14. 38.37Many VPCs: peering, Transit Gateway and the Reachability Analyzer
  15. 38.41Hybrid networks: Site-to-Site VPN, Direct Connect and PrivateLink
  16. 38.43Choosing the data and edge services, and what keeps them alive

27 hands-on labs (missions, incidents and drills) run in the terminal.

Open this chapter in the terminal Free, in your browser - a real Ubuntu terminal to try it in, with missions that check your work.

Questions people ask

Do I need the PostgreSQL chapter and AWS II before this one?

They help a lot. RDS runs the same PostgreSQL you operated by hand, so replicas, WAL and connection pools are familiar, and the labs reach private resources through Session Manager and VPCs that AWS II set up. Where a lesson needs one of those ideas, it says what it uses, so you can follow along without them.

Why are so many services in one chapter?

Because they answer the same few questions: where does the data live, which copy survives which failure, and how do users and other networks reach it. Seen side by side, RDS, DynamoDB and ElastiCache become a choice by access pattern, and Multi-AZ, point-in-time restore, replication and AWS Backup become a choice by the failure they cover. The last lesson puts those decisions in one place.

Will the labs cost money if I repeat them on a real account?

Some of them would, and by the hour whether traffic flows or not: RDS instances (twice for Multi-AZ), RDS Proxy, ElastiCache nodes, NAT gateways, Transit Gateway attachments and VPN connections. DynamoDB on-demand, S3 and CloudFront bill by use. If you practise on a real account, set a budget first, tag what you create and delete it the same day.

What is simulated and what is real in this chapter?

The commands, their options, the output formats and the error messages follow the real AWS CLI v2 and the current documentation. The account behind them is simulated (simulator): account 111122223333 in eu-central-1, with eu-west-1 as the DR Region. RDS and Aurora answer with a real PostgreSQL engine, but failovers take about 35 seconds instead of one to two minutes, and throttling and rate limits use lab-sized numbers. What you type works the same way against a real account.

Does this chapter help with the AWS certifications?

Yes. Every lesson ends with an "On the exam" note that maps the topic to SOA-C03, SAA-C03 and CLF-C02 domains and lists the classic question shapes, such as "survive an AZ failure with no data loss" or "serve a private bucket only through CloudFront". The labs give you the hands-on side that multiple-choice questions assume you have.