The primary Kubernetes event, KubeCon, just wrapped up in Barcelona. I watch the videos as they come out to learn more and keep up to date on happenings in that world. I’ll try to post to this thread what I watch and learn!
Fine grained permissions in Kuberneted
A lot of folks would like to give their JupyterHub users more rights on a Kubernetes cluster in a safe way. This talk is a nice high level overview of the general landscape here. It also pointed me to Open Policy Agent which sounds extremely interesting!
Please post your favorite videos too 
I think there are almost 3 kubecons a year, so there is tons of content,
3 Likes
Covers very basics of thinking about how to size your clusters. Complex topic, but this is a good 5min intro.
I’ve been interested in security a lot recently, partially because I’ve listened to over 40 hours of ‘true crime’ style podcasts about cybersecurity. Very scary world!
In kuberentes world, Falco seems to be the way to go. This talk has a nice overview and a cool demo. It will be interesting thinking of how this would work in a Jupyter context, since arbitrary code execution is sort of what we do. Nevertheless, very important and interesting.
Data Science workloads often vary in the amount of resources they need (CPU / RAM) over time. However, we usually size them in kubernetes to be the maximum CPU / RAM they need, since you can not change those values after the pod is created without a restart. Since notebooks and similar lose in-memory state when restarted, we try to never do those. This talk is about the massive changes coming to make this possible, focusing on batch jobs / notebook style pods
This is exciting, since it will let us dynamically change each user’s notebook resource requests based on historical resource usage.
Z2jh is based on Helm 2. Version 2 has some challenges - especially around the tiller component. Helm 3 addresses a lot of these, and there is an alpha just out to play with. This talk gives you a more general outline and a cool demo