Skip to content
Twitter Snowflake Server implemented in Python and Tornado for HTTP based ID generation
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.
.gitignore Initial commit Aug 30, 2012


pysnowflake is a Python implementation of Twitter's snowflake service -

This is based on the pysnowflak python trift service that Erans did -

Our use case wasn't thrift oriented but HTTP based, so this is a service that runs under the Tornado web framework. Which makes it possible to only have one depenancy (tornado).

Due to various reasons this implementation does not reach the performance indicated in the original Snowflake implementation, however it is good enough in most cases and can be combined with the help of a software load balancer such as HAProxy to run multiple processes to get higher performance.


  • Install Tornado
  • Run the service


usage: [--debug] [--port=9000] [--datacenter=DC_ID] [--worker=WORKER_ID]

Python based Snowflake server over HTTP

See the original snowflake server docs for a detailed description, but the bottom line is unique numbers will insure unique identifers generated. WORKER_ID is the identifier for this process DATACENTER_ID is the identifier for this datacenter

API (aka URLs):

/id/<USERAGENT>    -- get a unique ID, <USERAGENT> is provided for metrics purposes
/timestamp/        -- get the current timestamp for this host
/datacenter/       -- get the data center identifier for this process
/worker/           -- get the data center identifier for this process


Please report any issues via github issues

You can’t perform that action at this time.