Stabilize UAT API pod selection

.gitea/scripts/deploy-environment.sh

@@ -72,16 +72,32 @@ kubectl rollout status deployment/step3-api -n "${KUBE_NAMESPACE}" --timeout=240
 kubectl get --raw "/api/v1/namespaces/${KUBE_NAMESPACE}/services/step3-api:http/proxy/readyz" >/dev/null
 
 if [ "${DEPLOYMENT_ENVIRONMENT}" = "uat" ]; then
-  API_POD="$(kubectl get pods -n "${KUBE_NAMESPACE}" -l app=step3-api \
+  # A completed old replica can briefly remain visible immediately after a
-    --field-selector=status.phase=Running -o jsonpath='{.items[0].metadata.name}')"
+  # rollout. Select the newest pod that is both Running and container-ready
-  if [ -z "${API_POD}" ]; then
+  # instead of relying on the first item returned by Kubernetes.
-    echo "No running API pod is available for dynamic question promotion." >&2
+  PROMOTION_SUCCEEDED=0
+  for attempt in 1 2 3; do
+    API_POD="$(kubectl get pods -n "${KUBE_NAMESPACE}" -l app=step3-api \
+      --field-selector=status.phase=Running \
+      --sort-by=.metadata.creationTimestamp \
+      -o jsonpath='{range .items[*]}{.metadata.name}{"\t"}{.status.containerStatuses[0].ready}{"\n"}{end}' \
+      | awk '$2 == "true" { name = $1 } END { print name }')"
+    if [ -n "${API_POD}" ] \
+      && kubectl wait -n "${KUBE_NAMESPACE}" --for=condition=Ready "pod/${API_POD}" --timeout=60s \
+      && kubectl exec -n "${KUBE_NAMESPACE}" "${API_POD}" -- \
+        dotnet QuestionsApi.dll promote-dynamic-questions \
+          --source /app/content/dynamic_questions \
+          --apply; then
+      PROMOTION_SUCCEEDED=1
+      break
+    fi
+    echo "Dynamic question promotion attempt ${attempt} did not find a stable ready API pod; retrying." >&2
+    sleep 2
+  done
+  if [ "${PROMOTION_SUCCEEDED}" -ne 1 ]; then
+    echo "Dynamic question promotion failed after 3 attempts." >&2
     exit 1
   fi
-  kubectl exec -n "${KUBE_NAMESPACE}" "${API_POD}" -- \
-    dotnet QuestionsApi.dll promote-dynamic-questions \
-      --source /app/content/dynamic_questions \
-      --apply
 fi
 
 kubectl rollout status deployment/step3-web -n "${KUBE_NAMESPACE}" --timeout=240s