dashboard-ux-breakout

Kubernetes Dashboard UX breakout session 12.5.17, led by Rahul Dhide (rahuldhide) and Dan Romlein

  • Resources:

  • Notes

    • 2018 Dashboard strategy.

    • Kubectl access via Dashboard.

      • Dhaval: Provide context to issues.
    • Third-party Widgets.

    • Custom Views.

      • Dhaval: Custom views will be very useful to share the event details, contextual information, and logs with specific time ranges. It will help our development teams to quickly analyze the issues.

      • Henning: UI should allow users to pick the important properties in different views/widgets.

    • Integrations and CRDs.

    • Onboarding experience.

      • Jared: Currently there is no crosslinking between docs and the UI. We can explore the concept and impact further.
    • Design system.

      • e.g. Standard for how a pod’s status is displayed.
    • Different use cases/personas

      • application developer

      • application operator

      • (multi-)cluster operator

    • Feedback from Dhaval and Hennings:

      • “We don’t use Dashboard, because of lack of authorization control.”

      • “Dashboard today caters to the dev perspective, which is OK.”

        • But for ops, currently missing dependency stacks (e.g. OS version) “I want to know different node versions.”
      • “Our developers use Dashboard for the logs”

      • “When we perform troubleshooting & debugging we expect 30 min before and after incident.

      • Want to be able to link users to docs for more info; “What’s a pod? What’s a deployment?”

        • Contextual docs displayed in UI.

        • Idea: Dashboard could scrape docs.

      • Kubernetes docs working on expanding glossary

      • Dashboard is backwards compatible.

      • Demo of Kubernetes Operational View

      • Use Case:

        • Onboarding: explain resource limits vs. resource requests.

        • Quickly looking at a cluster and knowing what’s going on.

      • “I really like this UI”

      • Wanted to gamify K8s with this UI.

      • “My cluster has 118 nodes, so the ability to filter would be important.” (general scale issues)

      • “Problem we’re running into is that the number of nodes will crash the browser, so we need some way to select cluster first”

      • Defining a view is hard, Dashboard shouldn’t attempt to do that.

      • Use case for custom Dashboard re-skinning.

        • “My exec looks at the look and feel.”
  • Kubernetes Dashboard 2017 User Survey: https://docs.google.com/forms/d/e/1FAIpQLScnxeub_xh7Lp4iZO1RKdTgIYK_cTwqFKv1WD-Cue4tZHcbhw/viewform?usp=sf_link