K8 dot unames

[yhal-nesi]

Resolves the following issue #852 by sanitizing k8 namespace name. Deals with ".", "@", unicode, and converts upper case to lower case. Slug generator is based on JupyterHub code https://github.com/jupyterhub/kubespawner/blob/main/kubespawner/slugs.py and can also be used to sanitize other types of k8 objects.

The bootstrap script would also need to be modified to take advantage of that. https://github.com/OSC/ondemand/blob/master/hooks/k8s-bootstrap/k8s-bootstrap-ondemand.sh
Otherwise the change is not compatible.

Most of the work done by my colleague @DininduSenanayake.

[johrstrom]

I think we're overthinking this. Why not just gsub replace the bad characters?

The bootstrap script would also need to be modified to take advantage of that. https://github.com/OSC/ondemand/blob/master/hooks/k8s-bootstrap/k8s-bootstrap-ondemand.sh

Another reason to simplify this. Can this script compute the same hash/various other calculations? No, but it could tr or sed search and replace.

[yhal-nesi]

I guess I am just trying to make it work in all cases. Let's say we replace "." by "-", and then firstname.lastname is going to become firstname-lastname. But what if there is already a user with that username? Same with upper case to lower case conversion etc. It is a bit contrived and you could say that if someone has a directory like that it is their problem.

Another reason to simplify this. Can this script compute the same hash/various other calculations? No, but it could tr or sed search and replace.

The way I was testing this is to call ruby inside a script. This needs rb file in the same directory, but there might be a way to call a library too? Not very familiar with ruby.

export SANITIZED_USERNAME=$(ruby -r /opt/ood-hooks/k8s-bootstrap/slug_generator.rb -e "puts(SlugGenerator.safe_slug '$ONDEMAND_USERNAME')")
export NAMESPACE="${NAMESPACE_PREFIX}${SANITIZED_USERNAME}"

[johrstrom]

I guess I am just trying to make it work in all cases.

I'd rather have something simple that solves most cases we actually see in the wild. I think a simple gsub and tr would do this. Maybe we can query Etc for collisions?

For example, while € can be a legitimate username on some systems, we'd never actually see it in practice.

I commend you on your diligence to cover all cases, but as a maintainer I need to balance simplicity with the solutions the code provides. Which is to say, I'd rather cover most cases we'll actually see with simple solutions than to have a complex one that covers a lot of cases we'll never actually encounter. (not to mention the distribution and upgrade problems this would bring)