Skip to content

Instantly share code, notes, and snippets.

@tompave
Last active August 6, 2024 13:21
Show Gist options
  • Save tompave/8590031 to your computer and use it in GitHub Desktop.
Save tompave/8590031 to your computer and use it in GitHub Desktop.
commented nginx.conf for Ruby on Rails
# A commented nginx configuration file for Ruby on Rails
#
# Author: Tommaso Pavese
# tommaso@pavese.me
# http://tommaso.pavese.me
#
# License: http://www.wtfpl.net/
#
#
# Tested with:
# Ubuntu 12.04
# nginx 1.4.1 - 1.4.7
# Rails 3.2 - 4.0
# Unicorn, Thin and Puma
#
#
# Docs:
# http://nginx.org/en/docs/
#
# Configuration measurement units:
# http://nginx.org/en/docs/syntax.html
#
# Each module can define its variables, the core ones:
# http://nginx.org/en/docs/http/ngx_http_core_module.html#variables
#
#
# As reference, the official example for nginx + unicorn:
# https://github.com/defunkt/unicorn/blob/master/examples/nginx.conf
#
#
# In order to bind to port 80, the master process must be executed as root:
# $ sudo /usr/sbin/nginx -c /etc/nginx/nginx.conf
#
# How to execute the worker processes
user user_name group_name;
# A single worker is enough for load balancing and reverse proxing.
# However:
# Disk I/O can block an nginx worker (depends on disk read performance).
# If the server is serving a lot of static files (e.g. assets) it is
# a good idea to increase the number of worker processes.
worker_processes 1;
# The limit on the maximum number of open files for worker processes.
# "open files" is intended as UNIX open file descriptors.
# This overrides the limit set by the OS for the user the workers run as.
# Run `ulimit -a` in a shell to see the current limit.
worker_rlimit_nofile 2048;
error_log /var/log/nginx/error.log warn;
pid /var/run/nginx.pid;
# --------------------------------------------------------
events {
worker_connections 1024; # increase if you have lots of clients
accept_mutex off; # "on" if nginx worker_processes > 1
use epoll; # best for Linux 2.6+ ("kqueue" for FreeBSD, OSX)
}
# --------------------------------------------------------
http {
# disables emitting nginx version in error messages
# and in the “Server” response header field
server_tokens off;
# MIME types
include /etc/nginx/mime.types;
# fallback MIME type if the client doesn't specify it.
default_type application/octet-stream;
# log
access_log /var/log/nginx/access.log combined;
# TCP settings
tcp_nopush on;
tcp_nodelay off;
# use unix sendfile()
sendfile on;
# gzip compression
gzip on;
gzip_http_version 1.0;
gzip_proxied any;
gzip_vary on;
gzip_min_length 500;
gzip_disable "MSIE [1-6]\.";
gzip_types text/plain text/xml text/css
text/comma-separated-values
text/javascript application/x-javascript
application/javascript application/json
application/atom+xml;
# text/html is included by default
# According to the HTTP standard, headers with underscores are perfectly valid.
# However, nginx defaults to dropping headers containing underscores, as they
# might introduce ambiguities when mapping headers to CGI variables.
#
# Since this is not a problem with Ruby on Rails, we can safely enable them.
underscores_in_headers on;
# --------------------------------------------------------
# It's possible to organize the server configuration in different files.
#
# Typically, these site-specific files contain at least:
# - a server{} block
# - an upstream{} block
#
# The convention is to store configuration files in:
# /etc/nginx/sites-availbale/
#
# and then symlink them into:
# /etc/nginx/sites-enabled/
#
# with:
# $ ln -s /etc/nginx/sites-availbale/example.com.conf /etc/nginx/sites-enabled/example.com.conf
#
# After that, send a HUP signal to the nginx master to reload the configuration on the fly:
# $ sudo kill -HUP `cat /var/run/nginx.pid`
#
#
include /etc/nginx/sites-enabled/*;
# --------------------------------------------------------
# upstream destinations
#
# http://nginx.org/en/docs/http/ngx_http_upstream_module.html#upstream
#
# fail_timeout=0 means we always retry an upstream even if it failed
# to return a good HTTP response.
# This happens when the Unicorn master nukes a single worker for timing out.
#
# Sockets can be opened at any location where the user has writing access,
# even inside the rails_root/tmp/sockets directory
#
# We can also use IP addresses and domain names, and mix them together.
#
# If more than one destination server is specified, the requests
# are distributed using a weighted round-robin balancing method.
#
# Important:
# Unicorn workers can all listen on the same Unix socket (or TCP port)
# at the same time.
# However, this might not be true for other servers.
# A cluster of Thins, for example, will require a Unix socket per process.
# Look at the second upstream block for an example.
# This works well with Unicorn (yes, even with several workers)
upstream rails_app_one {
server unix:/tmp/rails_app_one.sock fail_timeout=0;
}
# This will work with a cluster of 3 Thin servers
upstream rails_app_two {
server unix:/tmp/rails_app_two.0.sock fail_timeout=0;
server unix:/tmp/rails_app_two.1.sock fail_timeout=0;
server unix:/tmp/rails_app_two.2.sock fail_timeout=0;
}
# This is weird, but it would still work.
upstream rails_app_three {
server unix:/tmp/rails_app_three.sock fail_timeout=0;
server 192.168.0.7:8080 fail_timeout=0;
server 127.0.0.1:3000 weight=3 fail_timeout=0;
server backendworker.example.com weight=5 fail_timeout=0;
}
# --------------------------------------------------------
# servers
# This server listens to port 80 and 443, and is configured to accept
# both HTTP and HTTPS traffic on the same domain.
# The Rails app can then decide to enforce HTTPS on the whole application
# or only on specific controllers.
#
server {
# Only one server can use the 'default_server' and 'deferred' options.
listen 80 default_server deferred;
# http://nginx.org/en/docs/http/ngx_http_core_module.html#server_name
#
# Also accepts lists of names, wildcard characters and regular expressions.
# Make sure that they all match the SSL/TSL certificate, if using one.
# See the last server block for more examples.
server_name www.example.com;
# HTTPS
# http://nginx.org/en/docs/http/configuring_https_servers.html
#
# If the key has a passphrase, nginx will require it each time it will
# boot or restart (making automatic management impossible).
listen 443 ssl;
ssl_certificate /etc/nginx/sslfiles/example_com_chained.crt;
ssl_certificate_key /etc/nginx/sslfiles/example_com_ssl_nopf.key;
ssl_session_cache shared:example_ssl_cache:1m;
ssl_session_timeout 5m;
client_max_body_size 4G; # default 1M
keepalive_timeout 20s; # default 75s
root /var/www/example/current/public;
# http://nginx.org/en/docs/http/ngx_http_core_module.html#try_files
#
# Tries URIs in sequence until one is found.
# First we try to serve static files directly from nginx.
# The rails application is the last alternative.
#
# See comment block at the end of the file for detailed examples.
#
# /maintenance.html can be symlinked into public/ to stop requests
# before they reach the rails app (e.g. while updating the DB).
# When it does not exist (normally), it's quickly skipped.
#
try_files /maintenance.html $uri $uri/index.html $uri.html @rails_app;
# The location directory allows very creative configurations.
# http://nginx.org/en/docs/http/ngx_http_core_module.html#location
#
# This is just a named location to be used in try_files.
location @rails_app {
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for; # the client IP
proxy_set_header X-Real-IP $remote_addr; # the client IP (again)
proxy_set_header X-Forwarded-Proto $scheme; # pass scheme (for HTTPS)
proxy_set_header Host $http_host; # the full host, for redirects within Rails
proxy_redirect off; # disable nginx redirect-rewrite logic
proxy_pass http://rails_app_one; # http is to be used even for https traffic
}
# Customize what html file should be served for different errors.
error_page 500 502 503 504 /500.html;
# This location block isn't strictly required, as '/500.html'
# can also be served by 'try_files $uri', above. Still, this
# location is more precise (thus has higher priority) and can
# be used to further customize the error response.
#
# location = /500.html { }
}
# A redirect from the naked domain to www
#
# This listens to port 80, thus it can't handle HTTPS traffic.
# (therefore, using $scheme instead of http is a bit useless)
#
# If the certificate is also compatible with the naked domain,
# then this server block can be improved by replicating here
# the SSL configuration of the main server.
server {
listen 80;
server_name example.com;
return 301 $scheme://www.example.com$request_uri;
}
# ----------------------------
# This server block features an efficient configuration
# to serve the pre-gzipped version of the static assets
server {
listen 80;
server_name .winter.is.coming.com
www.winteriscoming.com
*.winteriscoming.net;
client_max_body_size 4G;
keepalive_timeout 10s;
root /var/www/winteriscoming/current/public;
try_files $uri $uri/index.html $uri.html @rails_app;
error_page 500 502 503 504 /500.html;
location @rails_app {
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-Proto $scheme;
proxy_set_header Host $http_host;
proxy_redirect off;
proxy_pass http://rails_app_two;
}
# Location block for the assets.
# See: http://guides.rubyonrails.org/asset_pipeline.html#gzip-compression
#
# '^~ /assets/' is a 'prefix string', not a regex.
# '^~' will prevent further location searching after this location is matched.
#
# 'root' is inherited from the parent block. With the default Rails directory
# structure it should point to 'app_root/public'.
#
location ^~ /assets/ {
# To serve the pre-gzipped version of the files, if available.
# The non-compressed versions MUST be available too, or a 404 will be returned.
# The general gzip directives in the 'http' block are still applied.
gzip_static on;
# Since Rails automatically appends digest fingerprints to asset file names,
# we don't need to worry about expiring them.
expires max;
add_header Cache-Control public;
# Or:
# expires 12h;
# add_header Cache-Control "public, must-revalidate";
# The 'try_files' directive is NOT inherited from the parent block.
try_files $uri =404;
# We could also use a less strict directive that falls back to the Rails app
# if a requested asset cannot be found. In this case, the cache headers here
# configured will be ignored and Rails will set them automatically.
# try_files $uri @rails_app;
# Normally the Rails app takes care of serving the static file 'public/404.html'
# when a 404 occours. Here, however, it will be handled by nginx.
error_page 404 /404.html;
}
}
}
# A note on $uri and the try_files directive:
#
# try_files /maintenance.html $uri $uri/index.html $uri.html @rails_app;
#
#
# /maintenance.html
# if this file exists (likely a symlink), the app is probably not
# available (updating the DB?). We serve it directly.
#
# $uri
# it could be an asset.
# e.g. assets/application.css
#
# $uri/index.html
# it could be a directory in the file system. In that
# case we look for an index.html file.
#
# $uri.html
# it could be an actual html file.
# e.g. 500.html
#
# @rails_app
# finally, if all other tests fail, we pass the request to the app
#
@rmsy
Copy link

rmsy commented Jan 23, 2019

Very helpful, thank you!

@climatebrad
Copy link

How does this compare to the old capistrano-puma nginx template?

@tompave
Copy link
Author

tompave commented Aug 6, 2024

Hey @climatebrad, to be honest this nginx config is over 10 years old 😮.
I haven't relied on this set up in production for a long time, and I don't know how up to date it is, to be honest.
I suppose it might still be valuable as a starting point, but I would suggest to carefully review it before using it.

As to your question, I've never used capistrano-puma that I recall, but back in the day I remember reviewing a lot of similar set ups from across the internet and OSS Rails projects, when coming up with my config above. I did test it with Puma, but back then I was mainly running Unicorns.

@climatebrad
Copy link

I'm new to all of this and naively thought that DIY deploying a rails project to a clean Ubuntu VPS would be straightforward :) I'm updating a 2007 rails project, ha!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment