Перейти к содержимому

40M Rows, Zero Downtime: How Database Migrations Work

DevDecoded

0:00 / 0:00

40M Rows, Zero Downtime: How Database Migrations Work

35 просмотров · 9 дн. назад
DevDecoded
3 подписчика
35 просмотров · 9 дн. назад
How do you change a production database without taking it offline? This video shows the two ways a migration causes an outage, and the three techniques that avoid it: native online DDL, gh-ost, and expand-contract. Covers renaming a column without downtime, Postgres 11+ fast defaults, MySQL ALGORITHM=INSTANT, gh-ost vs pt-online-schema-change, and the Rails strong_migrations gem. CHAPTERS 0:00 Intro 1:11 Why migrations break production 4:13 The deploy / migration timeline 5:18 The landscape of techniques 6:11 Native online DDL (Postgres, MySQL) 8:15 gh-ost 11:07 Expand-contract (rename a column safely) 14:18 Comparing the three 14:50 The Rails code 17:01 Real-world usage 18:22 Challenges and traps 19:40 Mental model and recap WHO IT'S FOR Backend engineers who ship schema changes to a live Postgres or MySQL database. It assumes you know basic SQL and what a rolling deploy is. WHAT'S COVERED Lock contention outages, schema-code version skew, backfills, dual-write drift, and why rolling back a schema is harder than rolling back code. Follow DevDecoded for more system design, explained visually.