A small and fast HTTP based microservice for extremely minimal geoip location lookups.
It bundles into a ~5MB docker image that can serve over ~250K reqs/sec (uncached) from my laptop.
The API is intentionally extremely minimal and is designed to return only the absolutely most frequently needed geographic metadata for IP lookups.
The API has only one endpoint /
, and you just put the IP address (IPv4 and
IPv6 both accepted) directly in the URI path.
Example:
// $ curl http://${SERVER_IP}/89.160.20.112
{"country":{"iso_code":"SE"},"location":{"latitude":59.4333,"longitude":18.05,"accuracy_radius":200}}
Response reformatted for ease of human reading:
{
"country": {
"iso_code": "SE" // ISO 3166-1 country code
},
"location": {
"latitude": 59.4333, // Approximate latitude of IP
"longitude": 18.05, // Approximate longitude of IP
"accuracy_radius": 200 // Accuracy radius, in km, for the location
}
}
This package generally favors understandability of code over performance optimizations. That said, it is written in a way to be fairly highly performant, and combined with it's minimal nature, it can trivially handle a sustained 150,000 requests/second on my workstation. This actually makes it faster than any other similar off-the-shelf packages I tested in a quick informal survey. (Note: my benchmarking was intentionally not robust, and I'm certainly not trying to start any microbenchmark wars here.)
Compile with standard Go toolchain or download a binary for your platform from the Releases page.
Usage of tinygeoip:
-addr string
Address to listen for connections on (default ":9000")
-db string
Path for MaxMind database file (default "data/GeoLite2-City.mmdb")
-origin string
'Access-Control-Allow-Origin' header, empty disables (default "*")
-threads int
Number of threads to use, otherwise number of CPUs (default 8)
-verbose
Log all requests
You will need to provide a city-level precision GeoIP2 database file. Free GeoLite2 versions are available for download from MaxMindDB.
A Go library is provided for utilizing within native projects. A standard
http.Handler
interface is utilized for compatibility within standard Go HTTP
middleware setups. Alternately, if you are doing the lookups within your
existing application and not over the HTTP microservice, you can get an average
lookup result in approximately 1.2 microseconds.
For more information, see the GoDocs.
A docker image is automatically built from all tagged releases.
To utilize it, be sure to mount your MaxMindDB database as a volume so that the running container can access it.
[TODO: provide an example for folks not so familiar with Docker.]
🚧 The current API is considered unstable. This is just being
released and I'd like some feedback to make any potential changes before tagging
a v1.0
which will maintain API stability.
In other words, comments and feedback wanted!
-
klauspost/geoip-service
is where some of the initial inspiration for this was drawn. The primary difference is here we have a significantly more minimal API (with integration tests), which removes the need for caching or external JSON serialization libraries (things we initially had here but removed as the perf tradeoff was not benchmarking as significant versus the added complexity). I also wanted the API payload to be much smaller for client efficiency. -
bluesmoon/node-geoip
Seems well received, but uses "somewhere between 512MB and 2GB" of memory, which made it highly unsuitable for my purposes.
Software license TBD upon v1.0 release.
The chosen license should also contain an additional clause similar to:
"This software is not licensed for usage in any application related to censorship or preventing access to information based on geographic region. Legal action will be pursued against any entity who uses this software to knowingly violate this provision."