build installable file - OLPC XO laptop - Fedora 18

[tony37]

I would like to install jupyter notebook on the OLPC XO laptops. The software is based on Fedora 18 - Sugar version 0.110. It has Python 2.7 installed, but no Python 3.The XO has a file system limit of 1GB for the XO-1, 2GB for the XO-1.5, and 4GB for the XO-1.75 and XO-4. The latter two have ARM processors. 

Since a school has many laptops (200 is typical), it is not feasible to install by downloading from the internet for each install. What is needed is an installable image - tar or rpm. The recommended Anaconda strategy is not feasible due to these storage limitations. Pip uses /tmp to cache the downloaded tarballs. This could yield a set of tarballs to use in a local install; however, the XO /tmp has a limit of 50MB. The installed version of pip does not have a cache-dir option.

JupyterHub might be an option except there is no internet access at the schools so a local server is used. I believe having many laptops using Jupyter Notebook on the server would be too much for the processor. Our general strategy is to have processing done so far as possible on the individual laptops.

I am hoping someone will be able to guide me to a method to install Jupyter Notebook in this environment.

Tony

[Roland Weber]

Hello Tony,

have a look at Miniconda. It has about 35 MB, instead of the 500+MB of a full Anaconda installer. Then create an Anaconda environment that has only the packages relevant for your environment:

1. Miniconda

2. Jupyter Notebook and dependencies

3. additional science packages

Then create an installer image just with these packages. Miniconda should be able to install everything without an internet connection, if you have the relevant packages and the metadata about them in your installer image.

hope that helps,

  Roland

[Tony Anderson]

Hi, Roland

I tried Miniconda. However, it includes binary objects which do not run on an Arm processor. It apparently installs Python 2.7 which is already installed on the system. I am not sure why either Python 2.7 or Jupyter Notebook would have binary dependencies - if so, then packages for each architecture will be needed.

I suspect I will have to do this the hard way. In the past with patience I have been able to download a tarball and then attempt to install it offline. This leads to messages identifying a dependency to download. Proceeding recursively should result in an installable set of tarballs.

Tony

--
You received this message because you are subscribed to a topic in the Google Groups "Project Jupyter" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/jupyter/D3G3ZCTHgVs/unsubscribe.
To unsubscribe from this group and all its topics, send an email to jupyter+u...@googlegroups.com.
To post to this group, send email to jup...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/jupyter/4613a51a-7ce4-4746-b31f-ce2bbaa21fa0%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Johannes Feist]

Just an idea - could you create a virtualenv on one machine, install everything there, and then just tar up that folder? I haven't tried it, but I guess virtualenv folders should be portable between identical systems (maybe even different systems for the same OS).

Best, 

Johannes 

You received this message because you are subscribed to the Google Groups "Project Jupyter" group.
To unsubscribe from this group and stop receiving emails from it, send an email to jupyter+u...@googlegroups.com.

To post to this group, send email to jup...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/jupyter/09229bb3-d9a5-aab1-940f-5c13769ec7a1%40usa.net.

For more options, visit https://groups.google.com/d/optout.

--

Johannes Feist
IFIMAC & Departamento de Física Teórica de la Materia Condensada
Universidad Autónoma de Madrid
johanne...@uam.es

[Thomas Kluyver]

You should be able to get most of the packages by running 'pip wheel notebook' on an internet-connected machine. That will save a load of .whl files in the current directory. Transfer that folder to the target machine, and run 'pip install path/to/folder/*.whl'.

The tricky bit is pyzmq, Jupyter's one binary dependency (besides Python itself). You'll need to figure out a way to make a compatible build of that for the ARM processors.