fix: wait for Rancher Flux resources before rollout

.gitea/workflows/deploy.yml

@@ -387,13 +387,63 @@ jobs:
           KUBECONFIG: outputs/kubeconfig
         run: |
           set -euo pipefail
+          wait_for_resource() {
+            local namespace="$1"
+            local resource="$2"
+            local timeout_seconds="$3"
+            local elapsed=0
+
+            until {
+              if [ -n "${namespace}" ]; then
+                kubectl -n "${namespace}" get "${resource}" >/dev/null 2>&1
+              else
+                kubectl get "${resource}" >/dev/null 2>&1
+              fi
+            }; do
+              if [ "${elapsed}" -ge "${timeout_seconds}" ]; then
+                echo "Timed out waiting for ${resource} to exist" >&2
+                kubectl -n flux-system get kustomizations,helmrepositories,helmcharts,helmreleases || true
+                exit 1
+              fi
+
+              sleep 10
+              elapsed=$((elapsed + 10))
+            done
+          }
+
+          reconcile_helmrelease() {
+            local release_name="$1"
+            local reconcile_at
+            reconcile_at="$(date +%s)"
+            kubectl -n flux-system annotate "helmrelease/${release_name}" \
+              reconcile.fluxcd.io/requestedAt="${reconcile_at}" \
+              reconcile.fluxcd.io/resetAt="${reconcile_at}" \
+              reconcile.fluxcd.io/forceAt="${reconcile_at}" \
+              --overwrite
+          }
+
           echo "Waiting for Rancher..."
+          wait_for_resource flux-system kustomization.kustomize.toolkit.fluxcd.io/addon-rancher 600
+          kubectl -n flux-system annotate kustomization/addon-rancher reconcile.fluxcd.io/requestedAt="$(date +%s)" --overwrite
+          wait_for_resource flux-system helmrelease.helm.toolkit.fluxcd.io/rancher 600
+          reconcile_helmrelease rancher
+          kubectl -n flux-system wait --for=condition=Ready helmrelease/rancher --timeout=1800s
+          wait_for_resource "" namespace/cattle-system 600
           kubectl -n cattle-system rollout status deployment/cattle-system-rancher --timeout=900s
           kubectl -n cattle-system rollout status deployment/rancher-webhook --timeout=900s
           kubectl -n cattle-system wait --for=condition=Ready issuer/cattle-system-rancher --timeout=900s
           kubectl -n cattle-system wait --for=condition=Ready certificate/tls-rancher-ingress --timeout=900s
 
           echo "Waiting for rancher-backup operator..."
+          wait_for_resource flux-system kustomization.kustomize.toolkit.fluxcd.io/addon-rancher-backup 600
+          kubectl -n flux-system annotate kustomization/addon-rancher-backup reconcile.fluxcd.io/requestedAt="$(date +%s)" --overwrite
+          wait_for_resource flux-system helmrelease.helm.toolkit.fluxcd.io/rancher-backup-crd 600
+          wait_for_resource flux-system helmrelease.helm.toolkit.fluxcd.io/rancher-backup 600
+          reconcile_helmrelease rancher-backup-crd
+          reconcile_helmrelease rancher-backup
+          kubectl -n flux-system wait --for=condition=Ready helmrelease/rancher-backup-crd --timeout=1200s
+          kubectl -n flux-system wait --for=condition=Ready helmrelease/rancher-backup --timeout=1200s
+          wait_for_resource "" namespace/cattle-resources-system 600
           kubectl -n cattle-resources-system rollout status deployment/rancher-backup --timeout=900s
 
       - name: Restore Rancher from latest B2 backup