Systems exhibition01 / Problem 02 / Architecture 03 / How it works 04 / Failure thinking 05 / Reflection Inspect the sourceDatabase Stress Testing Framework on GitHub
Engineering tool · Concurrent JDBC workloads
Database Stress Testing Framework
A configurable Spring Boot tool for applying concurrent load to JDBC databases and watching throughput, tail latency, and failures as they happen.Java 17Spring BootJDBCRedisWebSocketPrometheusGrafanaOpenTelemetry
worker 01worker 02worker 03worker 04worker 05worker 06worker 07worker 08worker 09
JDBC target
connectionsp50p95p99
The constraint.
Database performance regressions are hard to catch because you cannot compare runs reliably — different load, different data, no plan-level visibility.
How it is shaped.
Each test gets an isolated HikariCP pool and fixed worker executor. A metrics collector publishes one-second windows over WebSocket; results persist separately from the target database and export to JSON or PDF.
Decisions made explicit.
- Per-test pool isolation
- Strategy-based workload profiles
- Execution-plan capture per query
- Bounded final latency samples and graceful executor shutdown
Where it can break.
Tests move through a defined lifecycle and share a stop signal. Errors are counted, classified, and surfaced alongside latency rather than discarded.
What I learned.
A useful load tool explains the workload and its failure modes; a large throughput number alone says very little.