Open yarikoptic opened 5 years ago
thx for bringing this up @yarikoptic! Your suggestion and the underlying software actually addresses an important point that is mentioned in the workshop: how can containers itself be made more reproducible? I'll definitely add datalad-containers
to the workshop, which will also result in some docker
related tests. I'm slightly addressing this topic during the docker management session talking about tagging
, but this part is far from being a sufficient coverage of the problem at hand. That being said, would actually do a complete session on this, heavily and specifically utilizing datalad-containers
. Depending on the participants and their background/experience it might then be worth to include a small datalad
section as well to ensure the best possible understanding. @miykael and I thought for a while about setting up a short datalad
introduction combining casts and slides. Would it be okay to ping you regarding both endeavours?
Of cause!
and forgot to mention another (sub)project I think you might find useful/relevant: https://github.com/ReproNim/containers/ . At large it is just a datalad dataset with containers pre-registered for datalad containers-run
.
ATM it has only a selected set of bids-apps Singularity containers, but may be I should finally try docker (and investigate docker-to-singularity proxy approach I mentioned in another issue - https://github.com/ReproNim/containers/issues/3), and may be just add "original" docker images and provide a helper also to run them via datalad containers-run
... not sure if I find time soon for that, so if you get interested -- please feel welcome to try/contribute.
https://github.com/datalad/datalad-container
Although used/tested in real life with singularity containers, there is preliminary proof of concept support for docker. This way you can keep your docker image under git annex to guarantee their authenticity and future availability (what if hub dies, or image gets replaced), while datalad containers-run would take care about instantiating the container from image within your docker service.