How we configure huge pages in ClickHouse Managed Postgres

Summary

ClickHouse Managed Postgres configures Linux huge pages to reduce the per-connection page-table cost and translation overhead of a large PostgreSQL shared_buffers cache. The configuration reserves a quarter of machine memory before PostgreSQL starts, verifies HugePages_Total, sets huge_pages = 'on' to prevent silent fallback, and sizes shared_buffers against the complete shared-memory segment, including buffer descriptors, WAL buffers, and lock tables. On an r7i.4xlarge with 32GB shared_buffers, 4KB pages added 31.1MB of page tables per connection and reached 6.12GB at 200 connections, while 2MB huge pages reached 111MB. A select-only pgbench test at 100 clients also rose from 373,083 TPS to 418,087 TPS, a 12% increase, while startup failure without a sufficient pool made misconfiguration explicit.

Takeaways

1. 1

   At 200 connections on the test system, page tables reached 6.12GB with 4KB pages but only 111MB with 2MB huge pages.

2. 2

   Huge pages must be reserved before memory fragmentation makes contiguous 2MB allocation unreliable; the allocation is checked through HugePages_Total.

3. 3

   The full shared-memory segment exceeds shared_buffers because it includes buffer descriptors, WAL buffers, and lock tables, so its size is measured with the Postgres binary before final sizing.