Conquering a Kubeflow Kubernetes Cluster with ksonnet, Ark, & Sonobuoy
summary
Demo shows Kubeflow deployed with ksonnet, validated with Sonobuoy (conformance run 30–60 minutes that surfaces multiple failing e2e tests), and migrated with Ark by backing up an entire Kubeflow namespace in AWS and restoring it to GKE (including TPU-enabled nodes). The demo scaffolds a 13+ microservice ML stack, generates and applies YAML via ksonnet, port-forwards a Jupyter notebook, and performs a full delete-and-restore—restore completes and notebook comes back up, proving cross-cloud portability end-to-end.
AI-written summary · about
chapters (7)
- 0:00Start
- 1:52And Not Just One Cloud!
- 2:59Kubernetes Helps!
- 5:56Heptio Has Answers
- 10:31ML Stacks are Composed LOTS of Microservices
- 11:55Vertical Composability TOO!
- 14:00Kubeflow to (Some of) the Rescue
related talks
51mMachine Learning using Kubeflow and Kubernetes
21mIntroducing Kubeflow (w. Special Guests Tensorflow and Apache Spark)
31mKubeflow - Data Science on Steroids
46mDeploying scalable Machine Learning models with Kubernetes
54mScaling Machine Learning Pipelines in Cloud
29mKubeflow: TensorFlow on Kubernetes
22mBuilding a MLOps Platform with Kubeflow on GKE
34mKubeflow: Multi-Tenant, Self-Serve, Accelerated Platform for Practitioners