A Linux 7.0 update caused PostgreSQL to slow dramatically on high-end hardware, dropping throughput by nearly half on a 96-vCPU machine. The shift stemmed from a change in the scheduler’s preemption rules, affecting how the kernel handles memory page switches. This adjustment, tied to how memory pages are managed, directly impacted database performance under heavy load.

Engineers observed a major bottleneck in buffer handling, highlighting the importance of kernel tuning for scalable databases. Understanding these details helps teams anticipate stability issues in production environments.