How to install nginx-mod in CentOS/RHEL 9 (aarch64)

High-performance web server

License: BSD Vendor: GetPageSpeed LLC https://www.getpagespeed.com/server-setup/nginx-mod-a-better-faster-nginx-build

Install

sudo dnf -y install https://extras.getpagespeed.com/release-latest.rpm
sudo dnf -y install nginx-mod

Description

nginx [engine x] is an HTTP and reverse proxy server, as well as a mail proxy server. This version adds some patches to improve performance: * HPACK by Cloudflare * Dynamic TLS records Built with QuicTLS to support HTTP/3, ensuring you're at the forefront of web technology. Be careful using this in production. This package merely addresses all the crazy folks who want you to compile NGINX to get the above features. The compiled NGINX is evil for many reasons. For more info, see: https://www.getpagespeed.com/server-setup/where-compilation-went-wrong The lesser, as well as faster evil is here. Packaged install is easier to use in production and is also easier to roll back to stable NGINX version in case there's an issue with those patches.

RPMs

nginx-mod-1.30.0-44.el9.gps.aarch64 [1.1 MiB]

Danila Vershinin (2026-04-23) — - slow_start.patch: wire the upstream ramp-up into the round-robin scheduler. · Previously the patch only parsed slow_start= and propagated it to the peer · struct so state files round-tripped cleanly, but ngx_http_upstream_get_peer · never scaled effective_weight, so freshly-live peers received their full · share from the first request. Now the scheduler uses peer->start_time as · the ramp-up anchor (an ABI-safe stub field already declared in upstream · nginx) and attenuates effective_weight by a 0..100 factor over the · configured window. Also fixes a pre-existing hunk mis-application that · double-assigned slow_start on the backup-zone path while missing the · backup-nonzone path. Also stamps start_time in the api module's POST/PATCH · paths so API-created peers ramp correctly. Verified on Rocky Linux 10 · with 2-peer and 3-peer upstreams (30s and 60s windows tracked linear · theoretical ramp within 0.5% at every sample point).