Jupyterhub + Sparkmagic/levy + Spark on K8s?

[neerawrt]

Does anyone have Jupyterhub + Sparkmagic/levy + Spark running on Kubernetes at scale (in production)?

How stable is it?

[itamarst]

I have no personal experience, but apparently Amazon EMR does Sparkmagic: https://aws.amazon.com/blogs/machine-learning/build-amazon-sagemaker-notebooks-backed-by-spark-in-amazon-emr/

[ogidogi]

Rather fine, but have some minor issues. Already 3 prod deployments.
Based on this version: apache/incubator-livy#167
Plus some small patches.

[itamarst]

@ogidogi if you have any useful patches to Sparkmagic, could you submit them as PRs? Thanks!

[jahstreet]

Feel free to refer the Helm setup ready for use: https://github.com/jahstreet/spark-on-kubernetes-helm . I'm opened for questions and feedback on usage. Can assist you with deployment.

[ogidogi]

Also some information can be found here: apache/incubator-livy#249

[PedroRossi]

@neerawrt I am currently using this exact architecture and the only problem I have found is that sparkmagic suport for jupyterlab has some bugs on the progress bar and other visual issues using the scala kernel

[wjxiz1992]

Hi @PedroRossi , is there any luck for us to see the sharing of how this architecture is set up on k8s?

[devstein]

Hi @wjxiz1992 there a some resources online for Sparks on K8s + Apache Livy. I know Viaduct has some open source repos for their docker images: https://github.com/viaduct-ai

[Wh1isper]

Hey guys, I've developed a module for deploying the latest 3.4.0 server-client mode on k8s and support config PySpark Session for direct connections!

https://github.com/Wh1isper/sparglim#spark-connect-server-on-k8s

Ping @devstein @wjxiz1992 @PedroRossi @jahstreet @ogidogi @itamarst @ogidogi @neerawrt