I'll chime in on this, too - We run Jupyter kernels inside Docker containers, running on an Apache Mesos cluster. This is used for online publishing <http://www.oreilly.com/oriole/index.html>.
Spark and related projects are popular among our authors at O'Reilly, so we've got a big need to be using Toree. Works great when run on laptops, though with an Apache 2.x version of Toree inside a container we encounter errors -- as Eszti described above, Some of our ops team at O'Reilly can rewrite the build, to work around these runtime errors in Docker. We'd like to contribute that back to the Toree project, rather than maintain separately. First we wanted to check with Toree devs, to see if that fits, doesn't conflict with your plans, etc. ? Also, have there been any related bug reports -- might be useful. Thanks, Paco PS: we may have some history here, not 100% sure: saw the commit from Osh -- over a year ago we had Jeremy Freeman's team (mostly Osh <https://github.com/andrewosh>) consult to develop a tutorial based on Spark in a container, and I have a hunch some of that code may have gone into Toree?
