EO Engineering Learning OS
PROJECT LADDER · P10
Back to map ↑
CORE PROJECT · TARGET L4 · 7 days
P10

High availability

Remove single-instance assumptions and design around failure domains.

01 / CONCEPT MAP

Know what the parts mean.

Learn each concept deeply enough to recognize it, place it in the system and reason about normal and failed behavior.

01

ALB

Define it · locate it · observe it · explain its failure mode.

02

Health checks

Define it · locate it · observe it · explain its failure mode.

03

Auto Scaling

Define it · locate it · observe it · explain its failure mode.

04

Stateless application

Define it · locate it · observe it · explain its failure mode.

05

Multi-AZ RDS

Define it · locate it · observe it · explain its failure mode.

06

Capacity

Define it · locate it · observe it · explain its failure mode.

02 / SYSTEM FLOW

Trace it end to end.

DNS→ALB→healthy targets across AZs→RDS Multi-AZ→S3

For every transition: identify input, output, identity, protocol, state, trust boundary, evidence and owner.

03 / GUIDED BUILD

Build, observe and explain.

Documentation and AI are allowed. The engineer must review every output and demonstrate the result from direct evidence.

  1. 01Create image or launch template
  2. 02Deploy across AZs
  3. 03Configure health checks
  4. 04Load test scaling
  5. 05Terminate one target
  6. 06Observe database failover behavior
04 / INCIDENT

All targets become unhealthy during deployment due to an incorrect health-check contract.

Required investigation

State impact → collect evidence → form competing hypotheses → test the cheapest discriminator → isolate root cause → contain → correct → verify.

Evidence pack

Timeline, relevant logs/metrics, failed assumptions, root cause, correction, verification and one prevention action.

AI ownership

AI may suggest causes and commands. The engineer must explain why each check is safe, what result is expected and how the result changes the hypothesis.

05 / DESIGN CHALLENGE

Defend the decision.

State availability objective, failure domains, capacity assumptions and cost trade-offs.

06 / VERIFY, SUBMIT & REVIEW

Submission evidence

Architecture review

  • Explain high availability without relying on memorized commands.
  • Draw the flow and name what crosses every arrow.
  • Identify the most likely, highest-impact and hardest-to-detect failure.
  • Show the evidence that proves the solution works.
  • Defend one security, reliability and cost trade-off.

Definition of Done

  • Acceptance criteria pass
  • Flow is drawn and explained
  • Security implications considered
  • Logs/metrics checked
  • Failure is tested
  • AI output is understood
  • Runbook is reusable
  • Mentor review passes
07 / MENTOR GUIDE

Do not score memory. Score the engineer’s ability to form a model, collect evidence, make a safe change and defend the trade-off.

1 Cannot explain2 Understands with gaps3 Implements and troubleshoots4 Designs and reviews