willianpc / freegeoip

Web server behind freegeoip.net

freegeoip.net

freegeoip.net is a public web service for searching geolocation of IP addresses. This is the source code of freegeoip.net's web server and script for building the IP database.

Overview

freegeoip.net is the result of a web server research project that started in 2009 hosted at Google's App Engine, using the Python API. A year later it moved to its own server infrastructure built on the Cyclone web framework, backed by Twisted and PyPy.

The current version is written in Go as the experiments progress with go-web and go-redis.

Install

List of prerequisites for building and running the server:

• Go compiler - for freegeoip.go
• libsqlite3-dev, gcc or llvm - for dependency go-sqlite3
• Python - for the updatedb script
• Redis - (optional) for API usage quotas
• The IP database

The following instructions are for Debian and Ubuntu servers.

Make sure Go is installed and both $GOROOT and $GOPATH are set, then run:

apt-get install build-essential libsqlite3-dev pkg-config
go get github.com/fiorix/freegeoip
cd $GOPATH/src/github.com/fiorix/freegeoip
go build

On recent OSX you might have to set the CC=clang before go build if the sqlite3 package fails to compile.

Proceed to building the IP database before starting the server.

Building the IP database

The IP database is composed of multiple files from multiple sources. It's a combination of IP subnets, country codes, city names, etc.

There's a helper script under the db directory that automates the process of building the database, and can be used regularly to update it as well.

It's a Python script called updatedb that creates ipdb.sqlite:

$ cd db
$ ./updatedb
... will download files and process them to create ipdb.sqlite
$ file ipdb.sqlite
ipdb.sqlite: SQLite 3.x database

This service includes GeoLite data created by MaxMind, available from maxmind.com.

Running

The server looks for freegeoip.conf in the current directory, but an alternative config can be specified using the -config command line option.

If the server is proxied by Nginx or another HTTP load balancer, edit the configuration file and set xheaders="true" and it'll use X-Real-IP or X-Forwarded-For HTTP headers (when available) as the client IP.

Run the server:

./freegeoip [-config /path/to/freegeoip.conf]

Then point the browser to http://localhost:8080.

If the IP database is unavailable (e.g. file does not exist, bad permissions) or redis is unreachable (if using redis as the quota backend), all queries will result in HTTP 503 (Service Unavailable).

We recommend supervisor for running the server in production. On Ubuntu, install apt-get install supervisor and drop the following config in /etc/supervisor/conf.d/freegeoip.conf:

[program:freegeoip]
user=www-data
redirect_stderr=true
directory=/opt/freegeoip
command=/opt/freegeoip/freegeoip
stdout_logfile=/var/log/freegeoip.log
stdout_logfile_maxbytes=50MB
stdout_logfile_backups=20

For listening on low ports as non-root user (e.g. www-data) on linux, set file capabilities at least once before running it:

/sbin/setcap 'cap_net_bind_service=+ep' /opt/freegeoip/freegeoip

Usage

Point the browser to http://localhost:8080 and search for IPs or hostnames.

Use curl from the command line to query the API:

$ curl -v http://localhost:8080/{format}/{ip_or_hostname}

It supports csv, json and xml as the output format. JSON supports callbacks with the callback query argument. The client (self) IP is used if ip_or_hostname is omitted in the query.

Examples:

$ curl -v http://localhost:8080/csv/
$ curl -v http://localhost:8080/xml/
$ curl -v http://localhost:8080/xml/freegeoip.net
$ curl -v http://localhost:8080/json/github.com?callback=foobar

If the server is listening on unix sockets, use nc to test:

echo -ne 'GET /json/my-domain.abc HTTP/1.0\r\n\r\n' | nc -U /tmp/freegeoip.sock

Credits

Thanks to (in no particular order):

• Gleicon for all the drama.
• Google for the map, Go, and AngularJS.
• Twitter for Bootstrap.
• MaxMind for the current database.
• ipinfodb.com for both the IP and timezones database back in 2010 and 2011.