Spring Boot 4.1 made virtual threads the default execution model, but this change can silently break HikariCP connection pooling. Under load, virtual threads hold database connections longer than platform threads, leading to pool exhaustion and request timeouts. This post dissects the problem: virtual threads are lightweight but can block on I/O without releasing connections promptly. The author provides concrete fixes including increasing pool size, reducing connection timeout, and using a separate thread pool for database operations. For teams migrating to Spring Boot 4.x, understanding this interaction is critical to avoid production outages. The post also discusses monitoring strategies to detect pool starvation early.
A deep dive into how Spring Boot 4.1's default virtual threads cause HikariCP pool exhaustion, with configuration workarounds.