PulseDB
PulseDB connects to PostgreSQL and MySQL instances and streams query plans, lock events, and replication lag into a unified dashboard. Engineers can drill into any slow query, see its EXPLAIN plan, and track improvements over time.
// key outcome
Helped users identify and fix P99 latency spikes in under 5 minutes vs. hours.
The Problem & User Context
Database performance issues are notoriously difficult to diagnose. DBA teams spent hours manually parsing pg_stat_statements and EXPLAIN output with no unified tooling for historical trends.
Approach & Tradeoffs
A lightweight Go agent runs beside each database instance, streaming pg_stat_activity and query plans to a ClickHouse time-series backend. The React dashboard renders flame-graph-style query waterfalls and sends Slack alerts on threshold breaches.
Technical Architecture
Go agent using database/sql with cgo-free drivers. ClickHouse for high-throughput metric ingestion. React + D3.js frontend. Prometheus exporter for scraping. Docker Compose for local setup, Helm chart for Kubernetes deployment.
// tech stack
// architecture diagram
Add your architecture diagram here (Excalidraw, Figma, or SVG)
Screenshots & UI States
Results & Lessons Learned
- Reduced MTTD for slow-query incidents from hours to under 5 minutes
- 1,800+ stars and 300+ production deployments
- Integrated with Datadog, PagerDuty, and OpsGenie
- Featured in PostgreSQL Weekly newsletter
Explore the Code
All source code is open-source on GitHub. Issues, PRs, and feedback are always welcome.