Containerized Clustered PostgreSQL
summary
Argues that containerized clustered PostgreSQL runs best on Kubernetes StatefulSet (petset) and shows working prototypes: a Patron/Patroni daemon run as PID 1 inside Postgres containers that registers roles in Kubernetes and uses etcd for leader election, producing automated failover in roughly 30 seconds. Demonstrates a Citus (cus) cluster: six data nodes with replication factor 2 plus a dedicated query node (seven nodes total) ingesting Wikipedia edits. Notes EBS persistent-volume bugs on AWS, points to Spilo/Patroni Helm charts and sidecar backup patterns for production.
AI-written summary · about
chapters (18)
- 0:00Start
- 0:34Code Examples
- 3:15Kubernetes
- 5:51Service Encapsulation
- 8:50Disadvantages
- 9:39Possible Performance Issues around Containerization
- 23:11Autopilot Pattern
- 24:27Service Discovery and Service Route
- 31:19Node Configuration
- 33:08Other Considerations
- 34:27Persistent Volumes
- 39:11Data Node Affinity
- 39:56Resources
- 42:14Why Wants a Separate Etz Cluster within Kubernetes
- 42:29Lesson One Is Performance
- 42:57Security
- 47:20Load Balancer
- 51:07How Does the Other System Make Sure that the Master Is Dead
related talks
44mPostgreSQL on PetSet
34mElephants on Automatic: HA Clustered PostgreSQL with Helm [I]
36mKube-native Postgres [I]
33mCitus & Patroni: The Key to Scalable and Fault-Tolerant PostgreSQL
26mKubernetes from the Database Out
50mPatroni Deployment Patterns
28mCitus on Kubernetes
38mPostgreSQL as a distributed computing platform