Closing the Observability Gap: Building a Kubeflow Plugin for Headlamp
summary
A Headlamp plugin built in TypeScript with the Headlamp Plugin SDK surfaces Kubeflow resources as tangible UI objects: five components, 16 CRDs, and 39 detail views and list views appear in a sidebar with status, timing, diffs and topology. The plugin auto-detects CRDs, shows notebook, Pipelines, Katib, Spark and training run details, offers a one‑click deep link to Kubeflow for authoring, and exposes the exact condition messages, resource requests, and logs needed to diagnose failures.
AI-written summary · about
chapters (19)
- 0:00Kubeflow Headlamp plugin intro
- 1:03Plugin architecture scope
- 2:30Plugin features and CRD detection
- 3:38Notebooks debugging overview
- 4:44Profile CRD namespace issues
- 5:27Kubeflow Pipelines overview
- 6:50Pipeline CRD diff view
- 7:34Run CRD SLA monitoring
- 8:16Experiment CRD model comparisons
- 8:58Katib trials component
- 10:05Trial CRD log analysis
- 10:48Suggestion CRD control loop
- 11:08Spark topology component
- 12:33Scheduled CRD concurrency issues
- 12:56Training job queue analysis
- 14:00Training job CRD conditions
- 14:41Runtime templates comparison
- 15:03CRD inspection and map
- 15:43Conclusion and resources
related talks
05mProject Lightning Talk: Let’s Talk About Headlamp, the Extensible Open-Source Kubernete...
23mYes, we need a new dashboard!
25mDoes Your Project Want a UI in Kubernetes-SIGs/headlamp?
18mKubeflow Ecosystem: Past, Present, Futu...
29mKubeflow Trainer Observability: Real-Time Progress Tracking for Reliab...
09mKubeflow Pipelines v2: The Next Generation of MLOps on Ku...
24mThe Good, the Bad, and the Missing Parts of Kubeflow
21mIntroducing Kubeflow (w. Special Guests Tensorflow and Apache Spark)