Stop Worrying and Keep Querying, Using Automated Multi-Region Disaster Recovery
Talkat Data on Kubernetes Day North America 2023 · on Monday, 2023-11-06· at 14:40· W471ab, McCormick Place·
summary
Automated orchestration uses Percona Operator for PostgreSQL and a multi-cluster control plane (Elotl Nova) to duplicate manifests across regions, wire the same S3/TLS secrets, and keep standby replicas in sync via streaming or WAL shipping. Failure detection hooks into monitoring, a failover controller runs a user-provided job that flips the standby to primary and reconfigures HAProxy, and a recorded demo with three AWS clusters shows a brief client glitch then seamless recovery — automation cuts manual steps and shortens RTO.
AI-written summary · about
chapters (21)
- 0:00Multi-region disaster recovery intro
- 1:02Why DR matters for databases
- 2:45Postgres operators basics
- 4:11DR options with Postgres operators
- 6:17Challenges in automated failover
- 7:41Myths about DR automation
- 8:46Introducing multi-cluster control plane
- 10:08What is a multi-cluster control plane
- 11:34Scheduling workloads across clusters
- 13:00Spread scheduling for DR and HA
- 14:24Disaster recovery workflow stages
- 16:12Failure detection and failover flexibility
- 17:35DR orchestration architecture
- 18:18Live demo setup
- 21:55Deploying Percona operator
- 22:38Configuring standby and primary
- 23:21Simulating region failure
- 24:45Automated failover demo
- 25:29Key takeaways and future work
- 26:12Resources and closing
- 26:32Q&A
related talks
30mComposing a Multi Database DBaaS from Open Source Kubernetes Operators
31mOpen Source Databases on Kubernetes: Best Practices
27mFrom Zero to Hero: Scaling Postgres in Kubernetes Using the Power of...
48mPostgreSQL in Kubernetes: Operation - Day 2
18mChaos testing of a Postgres cluster on Kubernetes
19mPreparing for disaster: going Multi-Region
41mCloud-Native Databases in Kubernetes with OpenShift and Postgres
50mDemystifying Kubernetes for Postgres DBAs; A Guide to Operators