Zero-downtime deployment on Kubernetes
summary
Claim: keep a Spring Boot e-commerce service live while changing a noncompatible DB schema by streaming live writes from the old to the new database. Implementation uses Hazelcast IMDG for in-memory state, Hazelcast Jet jobs plus a Debezium-style change‑data‑capture pipeline, MySQL targets, and Kubernetes rolling upgrades. Built a forward pipeline that copies/new‑formats cart rows so pods upgrade without blocking; demo mostly succeeds with a small live hiccup but no lost cart data.
AI-written summary · about
chapters (11)
- 0:00Introduction to zero downtime
- 3:10Blue-green deployments explained
- 8:05State management challenges
- 13:50Database migration strategies
- 19:04Change data capture pipeline
- 21:11Demo setup and initial state
- 24:25Preparing the green database
- 25:10Deploying the streaming pipeline
- 29:22Rolling upgrades to version 2
- 32:52Key takeaways and conclusion
- 34:58Q&A session
related talks
37mZero downtime Deployment on Kubernetes
35mZero-Downtime Deployment on Kubernetes
31mZero downtime deployment with Kubernetes, Spring Boot and Flyway
38mZero-downtime deployment with Kubernetes, Spring Boot and Flyway
43mZero-downtime deployment with Kubernetes, Spring Boot and Flyway
40mZero-Downtime Java Enterprise Applications With Kubernetes
40mStrategies for Zero Down Time, Frequent Deployments
25mAutomating Kubernetes Upgrades for Zero Downtime