Fix remove redundant component label added for backward compatibility with upgrade tests

deployments/helm/dra-driver-nvidia-gpu/templates/controller.yaml

@@ -35,7 +35,6 @@ spec:
       labels:
         {{- include "dra-driver-nvidia-gpu.templateLabels" . | nindent 8 }}
         {{- include "dra-driver-nvidia-gpu.selectorLabels" (dict "context" . "componentName" "controller") | nindent 8 }}
-        nvidia-dra-driver-gpu-component: controller
     spec:
       {{- if .Values.controller.priorityClassName }}
       priorityClassName: {{ .Values.controller.priorityClassName }}

deployments/helm/dra-driver-nvidia-gpu/templates/kubeletplugin.yaml

@@ -38,7 +38,6 @@ spec:
       labels:
         {{- include "dra-driver-nvidia-gpu.templateLabels" . | nindent 8 }}
         {{- include "dra-driver-nvidia-gpu.selectorLabels" (dict "context" . "componentName" "kubelet-plugin") | nindent 8 }}
-        nvidia-dra-driver-gpu-component: kubelet-plugin
     spec:
       {{- if .Values.kubeletPlugin.priorityClassName }}
       priorityClassName: {{ .Values.kubeletPlugin.priorityClassName }}

tests/bats/Makefile

@@ -20,8 +20,8 @@ TEST_CHART_REPO ?= "oci://gcr.io/k8s-staging-nvidia/charts/dra-driver-nvidia-gpu
 TEST_CHART_VERSION ?= "$(VERSION_STAGING_CHART)"
 
 # The baseline Helm chart to test upgrades from and downgrades to.
-TEST_CHART_LASTSTABLE_REPO ?= "oci://ghcr.io/nvidia/k8s-dra-driver-gpu"
-TEST_CHART_LASTSTABLE_VERSION ?= "25.12.0-0882da87-chart"
+TEST_CHART_LASTSTABLE_REPO ?= "oci://registry.k8s.io/dra-driver-nvidia/charts/dra-driver-nvidia-gpu"
+TEST_CHART_LASTSTABLE_VERSION ?= "0.4.0"
 
 # If not "false": the to-be-tested Helm chart is installed from the local
 # filesystem (from `deployments/helm/dra-driver-nvidia-gpu`). Make sure

tests/bats/helpers.sh

@@ -85,18 +85,16 @@ iupgrade_wait() {
   # not natively supported by `kubectl wait`, hence this must be something of
   # the shape
   # `kubectl get pods ... | xargs -I{} kubectl wait --for=condition=Ready pod/{} `
-  # TODO: change `nvidia-dra-driver-gpu-component` when last stable supports both, the
-  # new and the old label key
   sleep 1
-  kubectl wait --for=condition=READY pods -A -l nvidia-dra-driver-gpu-component=kubelet-plugin --timeout=15s
+  kubectl wait --for=condition=READY pods -A -l dra-driver-nvidia-gpu-component=kubelet-plugin --timeout=15s
 
   # Again, log current state.
   kubectl get pods -n dra-driver-nvidia-gpu
 
   # That one should be obvious now, but make that guarantee explicit for
   # consuming tests. Skip when compute domains are disabled (no controller deployment).
   if [ "${DISABLE_COMPUTE_DOMAINS:-}" != "true" ]; then
-    kubectl wait --for=condition=READY pods -A -l nvidia-dra-driver-gpu-component=controller --timeout=10s
+    kubectl wait --for=condition=READY pods -A -l dra-driver-nvidia-gpu-component=controller --timeout=10s
   fi
   # maybe: check version on labels (to confirm that we set labels correctly)
   log "iupgrade_wait: done"

tests/bats/test_gpu_updowngrade.bats

@@ -54,7 +54,7 @@ bats::on_failure() {
   _node=$(kubectl get pod "${_podname}" -o jsonpath='{.spec.nodeName}')
   log "workload runs on node: ${_node}"
   _kpod=$(kubectl get pods -n dra-driver-nvidia-gpu \
-    -l nvidia-dra-driver-gpu-component=kubelet-plugin \
+    -l dra-driver-nvidia-gpu-component=kubelet-plugin \
     --field-selector spec.nodeName="${_node}" \
     -o jsonpath='{.items[0].metadata.name}')
   log "kubelet-plugin pod on that node: ${_kpod}"
@@ -119,7 +119,7 @@ bats::on_failure() {
   _node=$(kubectl get pod "${_podname}" -o jsonpath='{.spec.nodeName}')
   log "workload runs on node: ${_node}"
   _kpod=$(kubectl get pods -n dra-driver-nvidia-gpu \
-    -l nvidia-dra-driver-gpu-component=kubelet-plugin \
+    -l dra-driver-nvidia-gpu-component=kubelet-plugin \
     --field-selector spec.nodeName="${_node}" \
     -o jsonpath='{.items[0].metadata.name}')
   log "kubelet-plugin pod on that node: ${_kpod}"
@@ -172,7 +172,7 @@ bats::on_failure() {
   # Stage 2: pick any kubelet plugin pod.
   local _kpod _node
   _kpod=$(kubectl get pods -n dra-driver-nvidia-gpu \
-    -l nvidia-dra-driver-gpu-component=kubelet-plugin \
+    -l dra-driver-nvidia-gpu-component=kubelet-plugin \
     -o jsonpath='{.items[0].metadata.name}')
   _node=$(kubectl get pod -n dra-driver-nvidia-gpu "${_kpod}" -o jsonpath='{.spec.nodeName}')
   log "targeting plugin pod ${_kpod} on node ${_node}"
@@ -200,7 +200,7 @@ bats::on_failure() {
   local _newkpod="" _deadline=$((SECONDS + 60))
   while true; do
     _newkpod=$(kubectl get pods -n dra-driver-nvidia-gpu \
-      -l nvidia-dra-driver-gpu-component=kubelet-plugin \
+      -l dra-driver-nvidia-gpu-component=kubelet-plugin \
       --field-selector spec.nodeName="${_node}" \
       -o jsonpath='{.items[0].metadata.name}' 2>/dev/null) || true
     if [ -n "${_newkpod}" ] && [ "${_newkpod}" != "${_kpod}" ]; then