Seesaw toolkit

An asynchronous toolkit for distributed web processing. Written in Python and named after its behavior, it supports concurrent downloads, uploads, etc.

This toolkit is well-known for Archive Team projects. It also powers the Archive Team warrior.

Installation

Requires Python 2 or 3.

Needs the Tornado library for event-driven I/O. The complete list of Python modules needed are listed in requirements.txt.

How to try it out

To run the example pipeline:

sudo pip install -r requirements.txt
./run-pipeline --help
./run-pipeline examples/example-pipeline.py someone

Point your browser to http://127.0.0.1:8001/.

You can also use run-pipeline2 or run-pipeline3 to be explicit for the Python version.

Overview

General idea: a set of Tasks that can be combined into a Pipeline that processes Items:

• An Item is a thing that needs to be downloaded (a user, for example). It has properties that are filled by the Tasks.
• A Task is a step in the download process: it takes an item, does something with it and passes it on. Example Tasks: getting an item name from the tracker, running a download script, rsyncing the result, notifying the tracker that it's done.
• A Pipeline represents a sequence of Tasks. To make a seesaw script for a new project you'd specify a new Pipeline.

A Task can work on multiple Items at a time (e.g., multiple Wget downloads). The concurrency can be limited by wrapping the task in a LimitConcurrency Task: this will queue the items and run them one-by-one (e.g., a single Rsync upload).

The Pipeline needs to be fed empty Item objects; by controlling the number of active Items you can limit the number of items. (For example, add a new item each time an item leaves the pipeline.)

With the ItemValue, ItemInterpolation and ConfigValue classes it is possible to pass item-specific arguments to the Task objects. The value of these objects will be re-evaluated for each item. Examples: a path name that depends on the item name, a configurable bandwidth limit, the number of concurrent downloads.

Consult the wiki for more information.