#High Availability Architecture

#Architecture Overview

The PostgreSQL Operator utilizes Patroni to achieve high availability, which primarily includes the following components:

1. Leader Election: The primary node is elected via etcd or the Kubernetes API.
2. Fault Detection: Continuous monitoring of the cluster's health status.
3. Automated Failover: Automatic switching when the primary node fails.
4. Configuration Management: Unified management of cluster configuration.

#Working Principle

1. Patroni monitors the status of PostgreSQL instances.
2. It regularly reports the status to the DCS (Distributed Configuration Store).
3. When the primary node fails:
   • The failure of the primary node is detected.
   • The failed node is removed from DCS.
   • A new primary node is elected.
   • The replica nodes are reconfigured.

#Configuration Parameters

Parameter	Default Value	Description
ttl	30	Primary node lease time (seconds)
loop_wait	10	Status check interval (seconds)
retry_timeout	10	Retry timeout (seconds)
maximum_lag_on_failover	1048576	Maximum allowable replication lag (bytes)

#Best Practices

1. Use an odd number of nodes (3 nodes recommended).
2. Configure reasonable resource limits.
3. Regularly test failover.
4. Monitor the cluster's health status.