I wrote a piece on the OpenEverest blog covering how to run high-availability PostgreSQL on Kubernetes using OpenEverest’s provider model.
It walks through deploying a three-node cluster with the Percona provider, verifying replication state across nodes, then two chaos tests: killing just the primary pod (which restarts on the same storage), and killing the pod, its persistent volume, and cordoning its node to force a genuine replica promotion — Patroni elects a new leader in about 1.4 seconds with zero data loss and updates endpoints automatically.
Read the full post on OpenEverest: High Availability PostgreSQL: From Zero to Cluster
Written as part of my ongoing contribution work to OpenEverest (CNCF Sandbox), where I’ve also opened a PR adding clickhouse-backup based backup/restore support for the ClickHouse provider — currently open for review.