Running PostgreSQL on Kubernetes in Production: Day 2 Operations with CloudNativePG
summary
CloudNativePG operator reliably provisions a primary plus read replicas and auto-creates secrets, but continuous backup not configured by default so Barman Cloud + S3 is essential for PITR; restores always create a new cluster (recovery is always a new cluster). Major Postgres upgrades trigger pg_upgrade on primary, wipe replicas and force full base backups to rebuild them, turning a 1 TB restore into hours (example: ~5.7 h at 25 MB/s; 3→5 replicas multiplies downtime), so measure IOPS, use a pooler, test drills and manage through GitOps.
AI-written summary · about
chapters (19)
- 0:00Introduction and speaker intro
- 1:36What is CloudNativePG
- 3:03Day 2 operations challenges
- 4:30Deploying a basic cluster
- 7:01Connecting via connection pooler
- 8:28Storage class and IOPS considerations
- 9:11Configuring backups with Barman Cloud
- 11:36Patching cluster for backup
- 12:43Scheduling automated backups
- 14:09Backup retention and recovery window
- 15:56Point-in-time recovery demo
- 17:46Restoring data to original cluster
- 18:49Recovery options and time to recovery
- 20:13Major version upgrades
- 21:42Downtime impact of major upgrades
- 23:50Topology and disaster recovery
- 25:38Replica clusters and continuous restore
- 27:26Key lessons and best practices
- 28:30Q&A session
related talks
34mRun PostgreSQL:The Kubernetes way
48mPostgreSQL in Kubernetes: Operation - Day 2
27mFrom Zero to Hero: Scaling Postgres in Kubernetes Using the Power of...
50mDemystifying Kubernetes for Postgres DBAs; A Guide to Operators
05mFive Minutes by CloudNativePG River
28mLightning Talk: Cloud Native PostgreSQL - Running PostgreSQL on Kubernetes
37mData On Kubernetes, Deploying And Running PostgreSQL And Patterns For Databases In a Kubernetes Cluster.
50mFrom VMs to Cloud-Native PostgreSQL in Kubernetes...