Track what keys are "done" using MySQL to implement a simple key-store.

mysql, keystore, batch, queue, LSF
pip install dones==0.2.0



The dones module can be used to mark whether a key is "done" and check whether a key has been marked "done". Keys can also be unmarked, so that they are no longer "done". Also, all keys can be unmarked by clearing the Dones. Keys are kept in their own namespace to avoid conflicts with other sets of other keys and to make it easy to implement clearing.

Why? I use dones to keep track of what I've already done. More specifically, I run large computations (a few million tasks) on a large cluster (a few thousand cores) with a slow filesystem (Isilon). When tasks inevitably fail, perhaps because the network storage goes offline, or a computer dies, or another user overwhelms the database with connections, I need to resubmit the tasks to the batch queuing system (LSF) of the cluster that are not already done.

The solution in this module fits my constraints. It handles the concurrent writes of a thousand jobs marking things done (not all at once). It is reasonably fast for reading and writing up to millions of jobs. This is important because or batch queue (LSF) only handles a few thousand jobs at a time gracefully. Finally, dones uses MySQL as a backend, which is important because I cannot run a key-value server like Redis on the cluster I use.


Feel free to make a pull request on github.


Awkwardly, dones is configured with a MySQL database url from the environment, so to test it, you need to add a url. For example:

DONES_DB_URL=mysql://myuser:password@localhost/mydb nosetests


  • Probably Python 2.7 (since that is the only version it has been tested with.)
  • MySQL-python PyPI package.


Install from

Download and install using pip:

pip install dones

Install from

Using github, one can clone and install a specific version of the package:

cd ~
git clone
cd dones
python install

Or use pip:

pip install git+git://


import dones

if not dones.get('my_pipeline').done('task1'):