pgbalancer 1.0 has been released, an AI-based PostgreSQL load balancer supporting REST API and MQTT.

The first release of pgbalancer, an AI-based PostgreSQL load balancer with REST API and MQTT, is now available. Pgbalancer provides connection pooling and load balancing for PostgreSQL, managed via HTTP protocol, and supports real-time event streaming. The project is a fork of pgpool-II with significant enhancements for intelligent request routing. The code is written in C and is distributed under the PostgreSQL open license.

Key features and capabilities:

• Connection pooling and load balancing.
• Automatic failover with watchdog support.
• Health monitoring and request caching.
• SSL/TLS, PAM/LDAP authentication.
• Compatibility with PostgreSQL 13-18.
• Part of the pgElephant utilities ensuring high availability. Can integrate with pgraft for Raft consensus, with monitoring supported by pgSentinel.
• AI-based load balancing.
  • Machine learning algorithms for intelligent request routing.
  • Training adapts based on response times and resource states. server.
  • Real-time state evaluation and request balancing predictions.
  • Configurable weight parameters for learning speed and randomness prediction speed.
  • Integration capability with monitoring platforms like Prometheus and Grafana.

  Management via REST API.

  • 17 HTTP/JSON functions for complete cluster management.
  • Integrated server API running on port 8080.
  • JWT HMAC-SHA256 token authentication.
  • Retrieving execution mode statistics and metrics.
  • Response time of less than 10 ms.

  MQTT event streaming.

  • MQTT is used to publish node state changes, failovers, and resource status checks in real time.
  • MQTT topics can be configured for different types of events.
  • Integration with Mosquitto, EMQX, or any MQTT broker.
  • Provides automatic notifications of abnormal events and offers a transport channel for cluster orchestration.
• Command-line utility bctl:
  • A unified interface replacing more than 10 separate 'pcp_*' commands.
  • 3 output formats: tables, JSON, or unchanged format.
  • Formatted table output with framing for runtime data pgbalancer.
  • Remote management and enhanced output mode.
• GitHub Actions Workflow:
  • build-matrix.yml: Builds for multiple platforms (Ubuntu, macOS, Rocky Linux).
  • Support for PostgreSQL 13-18.
  • DEB packages for Ubuntu 22.04/24.04, Debian 11/12.
  • RPM packages for Rocky 9, AlmaLinux 9, CentOS Stream 9.
  • Support for Autotools with autoreconf integration.
  • docs.yml: GitHub Pages documentation deployment.

Source: opennet.ru