Real-World Case Studies

Introduction

Lessons from real production incidents and migrations: GitLab's 2017 backup loss, Stripe's MongoDB→Postgres trade-offs, Shopify's MySQL sharding (Vitess).

Understanding the topic

Core concepts to understand:

• GitLab 2017: untested backups → near-total loss.
• Stripe: ACID-or-bust → Postgres for ledgers.
• Shopify + Vitess: horizontal scale on MySQL.
• Notion: scaling Postgres + sharding.
• Uber: schemaless on top of MySQL.

Syntax reference

Visual workflow / architecture:

Common patterns
 □ Backups untested = no backup
 □ Scale: replicate first, shard last
 □ Money belongs in ACID stores
 □ Migrations: dual-write, switch reads, retire old

Real-world use

Reading post-mortems is a free senior-engineer education. Imitate winners, learn from losers.

Best practices

• Read post-mortems regularly.
• Apply lessons before incidents.
• Discuss case studies in design reviews.

Hands-on exercise

Interview preparation — practice these questions:

• Q1. Lesson from GitLab 2017.
• Q2. Why did Stripe pick Postgres?
• Q3. How does Vitess work?
• Q4. Migration pattern: dual-write & switch.
• Q5. Schemaless on MySQL — pros/cons.