search

jupyter / docker-demo-images

Demo images for use in try.jupyter.org and tmpnb.org
BSD 3-Clause "New" or "Revised" License
75 stars 132 forks source link

container bloat for demo image #65

Closed kbroughton closed 9 years ago

kbroughton commented 9 years ago

The pull of the container from 2 weeks ago jumped from 2.5 Gb to 6 Gb.
Why is that? $ docker images | grep jup jupyter/demo latest d6449de4bebe 2 weeks ago 6.234 GB jupyter/notebook latest c9a1d70bfcdf 3 weeks ago 910.2 MB jupyter/minimal latest 9c79b1043d59 5 weeks ago 1.184 GB jupyter/demo clean 66f14ab192df 3 months ago 2.578 GB jupyter/demo-minimal latest 98b7aad170f6 9 months ago 791.2 MB

We are trying to customize the container based on jupyter/demo and the builds are failing on pushing to the registry with out of disk space errors.

rgbkrk commented 9 years ago

Yeah that's kind of weighty. If I had to guess, that's the new Spark stuff. /cc @parente

We haven't built an uploaded this one yet due to transitions to Jupyter 4.0

parente commented 9 years ago

Before I rebuilt the image with Scala + Spark, I measured it at ~5GB. I'll double-check if I can find the SHA.

At any rate, I agree the image is massive, but intentionally: it's meant to be a demo of the multi-language support of Jupyter. If smaller images are desired for other purposes, I think they belong in jupyter/docker-stacks.

parente commented 9 years ago

Commit ccfa2724, just before the Scala kernel and Spark support for Scala and Python3 went in builds as:

jupyter/demo latest edd16f44e83e 11 seconds ago 5.409 GB

After:

jupyter/demo stable e0b0cab62986 6 weeks ago 6.186 GB

parente commented 9 years ago

Is there any action to take on this item?

kbroughton commented 9 years ago

I don't think so. You might point in the Readme to the docker-stacks repo for anyone looking for a more minimal container docker-stacks/minimal-notebook

parente commented 9 years ago

Good idea. I was thinking about a FAQ in the README with other content as well. I'll work it in.

rgbkrk commented 9 years ago

One point I'm realizing out of this - with multiple tags, if people do docker pull jupyter/demo they're getting all these mega images.

parente commented 9 years ago

With older docker clients, yes. Modern docker (I forget which version) only pulls latest by default unless you specify --all on the pull.

Update: PR that changed the behavior in May, 2014 https://github.com/docker/docker/pull/5841

rgbkrk commented 9 years ago

ohhhh, neat

parente commented 9 years ago

I think we can close this. There's now a FAQ with pointers to docker-stacks.