Skip to content

Cluster Workloads Pods/Jobs — newest-first infinite scroll

Motivation

On a long-lived ElasticBLAST cluster the Workloads card's Pods and Jobs tabs rendered every row at once. A real cluster accumulated 500+ pods and 1000+ Jobs, so the tables painted hundreds of rows in creation order (oldest first), burying the rows an operator actually cares about — the most recent ones — at the bottom of a very long scroll.

User-facing change

  • The Pods and Jobs tabs of the cluster Workloads card now:
  • Sort newest-first by Kubernetes age (creationTimestamp) — the most recently created pod/job is at the top.
  • Render 20 rows at a time inside a fixed-height (≈460px) scroll viewport, growing the window by 20 as the operator scrolls (infinite scroll via IntersectionObserver).
  • Show a compact footer: Showing 20 / 506 · scroll for more, or 506 total once everything is rendered.
  • The window resets to the first 20 rows when the namespace filter changes or a refetch returns a different snapshot, so the operator is never parked deep in a stale tail.
  • Deployments tab is unchanged (its roster is small).
  • No backend, API, or payload change — this is purely client-side ordering and windowing. The namespace filter's "N shown" count is unchanged (it still reflects the total rows in the selected namespace).

Implementation summary

Validation evidence

  • cd web && npm run build → tsc -b && vite build ✓ built (no type errors).
  • npx eslint on the four touched/added files → clean (exit 0).
  • No ClusterDiagnostics unit tests exist; the panels' only consumer (K8sWorkloadsSection) passes unchanged props.