Share data files with your users#

Sometimes you might need to distribute a set of files to all your users, so they don’t have to re-download it once per person. This is particularly useful in educational contexts, where you might be teaching a course that reads a common dataset.


If you are teaching with large datasets, you might run out of memory! So consider teaching with just a subset of data before distributing large datasets to your users.

The shared directory#

There are two folders that are used together to allow Administrators to share data files with all users.


All users have a directory called shared in their home directory. This is a readonly directory - users and administrators can not write to it. However, anybody can access and read from the shared directory. This is how a user accesses a data file distributed by a hub administrator.


(administrators only) Admin users also have a directory called shared-readwrite in their home directory. This is the same folder as the shared directory, but writeable! Any files admins put here will be immediately visible in all users’ shared directories.

The allusers directory - available upon request#

Sometimes, hub Administrators might need to share data files with the users, and this files must only be visible to the users that they’re addressed to. For example, uploading graded notebook assignements in each user’s home directory.

If such a workflow is needed, then an additional allusers directory can be enabled for administrators only, where all the hub users’ directories are accessible to read and modify. Please reach out to us if you’d like this feature enabled.


Please keep in mind that enabling this feature, means that any admin user could access all users’ home directories, and possibly delete them by accident, if not careful. So, this feature should be used with extra caution!

A workflow for sharing datasets#

To share datasets with users, admins should put the dataset in ~/shared-readwrite. If they are distributing notebook / content that reads this dataset, it should refer to files in ~/shared/ rather than in ~/shared-readwrite. This will prevent accidental erasures / writes on behalf of admins.


This is an experimental feature, and the names of these directories and their structure are subject to change.