18 February 2016
Yesterday I did a talk about Openstack TaskFlow at the Amsterdam Python Meetup. TaskFlow is a cool Python library that can help you write code that can deal with unreliable systems (like flaky APIs with calls that need retrying and/or reverting). It also enables you to continue a flow (collection of tasks) where it left off if the original worker died with relative ease. With TaskFlow you can use a jobboard (either based on Zookeeper or Redis) which makes it very scalable.
Conductors atomically pick up jobs from the jobboard and record their state in a persistence backend (Zookeeper, MySQL or some other options), so at all times the system knows how far a flow has progressed and what actions still need to happen.
The slides were created by converting a Jupyter (previously named IPython) notebook to html with jupyter nbconvert:
jupyter nbconvert presentation.ipynb --to slides
And cloning the HTML presentation framework reveal.js into the same directory as the generated .html file.
git clone https://github.com/hakimel/reveal.js/