docker-stfd)docker-stfd is a tool which massively reduces the size of existing Docker images by running them, monitoring what files in the filesystem are actually being used, deleting the ones that aren't, then rebuilding the top image layers into one. It makes use of dockerfile-from-image.
This will only work in the case that the exact same set of files are accessed each time a container is run, otherwise docker-stfd will not know which are safe to delete. It will also leave your container in a highly volatile state, with anything other than the original command unlikely to work.
Nevertheless, this is useful for containers which perform a straighforward deterministic calculation each time, and where size if very important. For example, we use docker-stfd to slim down images before sending them to the volunteers at Cosmology@Home.
If you would normally run your container with,
docker run <image> <cmd>then,
docker-stfd <image> <cmd>creates a slimmed down version (with default name <image>-slim).
<cmd> contains the same option flags as docker-stfd, you must use -- to separate them, i.e. docker-stfd <image> -- <cmd>. timeout X <cmd> instead of <cmd>.$ ./docker-stfd myimage:latest timeout 10 python mycode.py
Reverse engineering Dockerfile...
Inspecting container file usage...
Getting image filesystem...
Getting base filesystem...
Creating new slimmed filesystem...
Building new container...
Created image myimage:latest-slim
Uncompressed size shrunk from 644.6 MB to 211.6 MBstrace and docker installed. find, grep, sort, and readlink. Most Debian/Ubuntu guests have this by default. This is not yet compatible with Busybox's grep. I have not tested others.