Skip to content

Cluster modal Workloads card — Pods / Deployments / Jobs tabs

Motivation

The cluster detail modal's Kubernetes diagnostics only listed Pods. The Azure portal Workloads pane shows Deployments, Jobs and more, and for an ElasticBLAST cluster the Job → Pod relationship is the natural unit of work: operators want to see finished/in-flight Jobs and the supporting Deployments (frontend, CoreDNS, metrics-server) without leaving the dashboard. A flat single list could not carry three resource shapes, so the Pods section was promoted to a tabbed Workloads card.

User-facing change

  • The cluster modal's "Pods" section is now a collapsible Workloads card with three tabs: Pods | Deployments | Jobs (the subset that matters for ElasticBLAST; the portal's Replica sets / Stateful sets / Daemon sets / Cron jobs were intentionally omitted).
  • Node-level diagnostics (Node Resources, Nodes) stay as sibling stacked sections — only the workload views are tabbed.
  • Each tab has the same namespace filter and "N shown" indicator. Tab labels show a live item count once loaded.
  • Pods — unchanged behaviour (all phases, Node / Pod IP columns, Logs / Describe / Delete actions).
  • Deployments — read-only: NS / NAME / READY / UP-TO-DATE / AVAILABLE / AGE. READY turns amber when ready < desired.
  • Jobs — read-only: NS / NAME / COMPLETIONS / STATUS / DURATION / AGE. STATUS is colour-coded (Complete / Failed / Running / Pending).
  • Tabs fetch lazily — opening the modal no longer fans out three Kubernetes API calls; a tab only loads when the card is expanded and the tab is active. "Refresh All" refetches node data and invalidates the live workload tab.

API / IaC diff summary

  • New routes (read-only, MSAL-gated, cluster-gated cache + _graceful degrade, same as /aks/pods):
  • GET /api/monitor/aks/deployments → { "deployments": K8sDeployment[] }
  • GET /api/monitor/aks/jobs → { "jobs": K8sJob[] }
  • New service helpers in api/services/k8s/monitoring.py (re-exported via the api/services/monitoring facade):
  • k8s_get_deployments(...) — apps/v1 list, ready/up_to_date/ available parsing, optional namespace scoping.
  • k8s_get_jobs(...) — batch/v1 list, completions and derived status (Complete/Failed/Running/Pending), start_time/completion_time.
  • No IaC change. No new Storage/network surface, no SAS, no shell-out — direct Kubernetes API via the existing kubeconfig token path.
  • Frontend: new K8sDeployment / K8sJob types + k8sDeployments / k8sJobs typed clients; new K8sWorkloadsSection (tab container) and K8sDeploymentsPanel / K8sJobsPanel / refactored K8sPodsPanel; shared useNamespaceFilter hook + NamespaceFilter component; formatDuration helper. Old K8sPodsSection deleted.

Validation evidence

  • uv run pytest -q api/tests/test_k8s_get_deployments_jobs.py api/tests/test_k8s_get_pods.py → 7 passed (new test covers endpoint URLs, namespace scoping, replica parsing, and all four derived Job statuses).
  • uv run pytest -q api/tests -k "aks or monitor or k8s" → 402 passed (no regression from the facade / route additions).
  • uv run ruff check on the touched backend files → All checks passed.
  • npx tsc --noEmit → no errors in any ClusterDiagnostics / monitoring.ts file (pre-existing unrelated SequenceDetail.tsx errors untouched).
  • npx eslint src/components/ClusterDiagnostics src/api/monitoring.ts → clean.