VectorAI DB Observability Stack

A production-ready observability stack for Actian VectorAI DB using Prometheus and Grafana.

What's included

• docker-compose.yml - Three-service stack: VectorAI DB, Prometheus, Grafana
• prometheus/prometheus.yml - Prometheus scrape configuration targeting VectorAI DB on port 6573
• prometheus/alert_rules.yml - Eight recommended alert rules from the VectorAI DB monitoring documentation
• grafana/dashboard.json - Importable Grafana dashboard with four panels: REST request rate, p95 latency, gRPC error ratio, and memory pressure
• scripts/load_test.py - Load test script: 1,000 random vector queries at 50 req/s using VectorAIClient

Prerequisites

• Docker and Docker Compose
• Python 3.10 or higher
• VectorAI DB community edition: https://www.actian.com/databases/vectorai-db/community-edition/
• actian-vectorai-client: pip install actian-vectorai-client

Quick start

docker compose up -d

Navigate to localhost:9090/targets to verify Prometheus is scraping VectorAI DB. Navigate to localhost:3000 to access Grafana and import the dashboard from grafana/dashboard.json.

Alert rules

Replace the placeholder values in prometheus/alert_rules.yml before deploying:

• Memory limit: update the value in VectorAIHighMemoryUsage to match your available RAM in bytes
• File descriptor limit: update the value in VectorAIFileDescriptorExhaustion to match your system fd limit