AWS engineer reports PostgreSQL perf halved by Linux 7.0, fix may not be easy (phoronix.com)

Article summary

An AWS engineer reported that PostgreSQL performance is halved on Linux 7.0, and a fix may not be easy. The issue is related to a configuration change in the Linux kernel that affects lock holder preemption. The problem may not affect all users, particularly those running PostgreSQL on smaller machines or with different configurations. A potential fix involves using the rseq slice extension.

Main themes

• PostgreSQL performance
• Linux kernel changes
• Database configuration
• Server optimization
• Upgrade strategies

What commenters say

• The performance regression in PostgreSQL on Linux 7.0 may not be universally experienced, and its impact depends on specific use cases and configurations.
• Using huge pages can significantly improve performance, but it may not always be the best approach, and its effectiveness depends on the specific workload and system setup.
• The decision to upgrade to new software versions should be based on careful consideration of potential benefits and risks, rather than simply following the latest releases.
• Some users argue that staying on currently-supported versions of software is important, while others believe that upgrading too quickly can lead to unnecessary problems and instability.
• The use of huge pages is not a simple system-wide toggle, but rather a complex configuration that requires careful consideration of various factors, including workload, system setup, and potential benefits.
• There are workloads where huge pages do not bring any measurable benefit, and the decision to use them should be based on careful evaluation of their effectiveness in specific contexts.