Repository files navigation

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 forArchive Team projects. It also powers theArchive Team warrior.

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.

To run the example pipeline:

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

Point your browser tohttp://127.0.0.1:8001/.

You can also userun-pipeline2 orrun-pipeline3 to be explicit for the Python version.

General idea: a set ofTasks that can be combined into aPipeline that processesItems:

• AnItem is a thing that needs to be downloaded (a user, for example). It has properties that are filled by theTasks.
• ATask 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.
• APipeline represents a sequence ofTasks. To make a seesaw script for a new project you'd specify a newPipeline.

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

ThePipeline needs to be fed emptyItem objects; by controlling the number of activeItems you can limit the number of items. (For example, add a new item each time an item leaves the pipeline.)

With theItemValue,ItemInterpolation andConfigValue classes it is possible to pass item-specific arguments to theTask 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.

Consultthe wiki for more information.