[onedr0p]

What

Replace the two per-container resource keys — gatus.resources and sidecar.resources — with a single top-level resources that maps to the Kubernetes pod-level resources field (Pod.spec.resources).

Why

This chart is a multi-container pod: the gatus container plus the gatus-sidecar native sidecar (an init container with restartPolicy: Always). Pod-level resources let you set one budget shared across both containers instead of splitting requests/limits per container and guessing the split. That's the natural model here now that Kubernetes supports it.

Behavior

• Rendered at the pod spec level, only when set ({{- with .Values.resources }}), so the default ({}) leaves the field unset.
• Not duplicated onto the containers — the gatus and sidecar containers no longer carry their own resources blocks.

# values.yaml
resources:
  limits:
    memory: 256Mi
  requests:
    cpu: 50m
    memory: 128Mi

renders as:

spec:
  template:
    spec:
      resources:        # <- pod level
        limits: { memory: 256Mi }
        requests: { cpu: 50m, memory: 128Mi }
      initContainers:
        - name: gatus-sidecar   # no resources block
      containers:
        - name: gatus           # no resources block

Compatibility

The chart now requires Kubernetes 1.34+ — kubeVersion is raised to >=1.34.0-0. That's where PodLevelResources is beta and on by default, so the top-level resources works without a feature gate, and Helm refuses to install on 1.29–1.33 rather than silently dropping the field.

The Helm e2e kind cluster moves to a 1.34 node image (kindest/node:v1.34.0, which needs the kind binary at v0.30.0+) so helm install satisfies the new constraint and actually exercises pod-level resources. The Go-binary e2e job is left unchanged — it tests the sidecar against the API server, not the chart, so it isn't gated by kubeVersion.

Migration (breaking)

This release requires Kubernetes 1.34+. And if you set either per-container resources key, move the values up to the top level:

-gatus:
-  resources:
-    requests: { cpu: 10m, memory: 64Mi }
-sidecar:
-  resources:
-    requests: { cpu: 10m, memory: 32Mi }
+resources:
+  requests: { cpu: 50m, memory: 128Mi }   # one shared pod budget

release-please will cut 0.3.0 (breaking change, bump-minor-pre-major).

Verification

• helm lint — pass
• helm unittest — 47 passed (added: pod-level resources rendered when set; omitted by default and never set per-container)
• helm template — confirmed resources renders at pod level with neither container carrying its own block
• kubeVersion enforced — helm template --kube-version 1.33.0 fails (chart requires kubeVersion: >=1.34.0-0 which is incompatible with Kubernetes v1.33.0); 1.34.0 renders
• README + values.schema.json regenerated (mise run generate); the gatus.resources/sidecar.resources rows are gone, a single top-level resources row/property is present, and the README requirements line now reads Kubernetes: >=1.34.0-0

---

Also in this PR — config.existingConfigMap is now templated

config.existingConfigMap is rendered through tpl, so the ConfigMap name can be a template expression (e.g. a release-derived {{ .Release.Name }}-config), matching the rest of the chart's tpl-rendered values. The required guard is preserved, so an unset value still errors clearly. Non-breaking; a plain string name behaves exactly as before. Covered by a new unittest (47 total) and confirmed by helm template.