Upload files to "root"

root/recover-stable-startup.sh

@@ -0,0 +1,58 @@
+#!/usr/bin/env bash
+set -Eeuo pipefail
+IFS=$'\n\t'
+
+CLUSTER_NAME="${CLUSTER_NAME:-lab-cluster}"
+RUNNER_NAME="${GITEA_RUNNER_CONTAINER:-gitea-runner}"
+KUBECONFIG_PATH="${KUBECONFIG_PATH:-/DATA/.kube/config}"
+
+echo "== Desactivando watchdog anterior =="
+sudo systemctl disable --now k3d-lab-ensure.timer 2>/dev/null || true
+sudo systemctl stop k3d-lab-ensure.service 2>/dev/null || true
+sudo systemctl reset-failed k3d-lab-ensure.service 2>/dev/null || true
+
+mapfile -t cluster_containers < <(
+  docker ps -aq --filter "label=k3d.cluster=${CLUSTER_NAME}"
+)
+
+if (( ${#cluster_containers[@]} == 0 )); then
+  echo "ERROR: no se encontraron contenedores del clúster ${CLUSTER_NAME}." >&2
+  exit 1
+fi
+
+echo "== Restaurando políticas confirmadas =="
+docker update --restart=unless-stopped "${cluster_containers[@]}" >/dev/null
+
+if docker inspect "$RUNNER_NAME" >/dev/null 2>&1; then
+  docker update --restart=unless-stopped "$RUNNER_NAME" >/dev/null
+  if [[ "$(docker inspect -f '{{.State.Running}}' "$RUNNER_NAME")" != "true" ]]; then
+    docker start "$RUNNER_NAME" >/dev/null
+  fi
+fi
+
+echo "== Recuperando k3d sin detener nodos activos =="
+if ! k3d cluster start "$CLUSTER_NAME" --wait --timeout 360s; then
+  echo "AVISO: k3d devolvió error; se esperará la API porque K3s puede continuar iniciando." >&2
+fi
+
+export KUBECONFIG="$KUBECONFIG_PATH"
+
+deadline=$((SECONDS + 360))
+while (( SECONDS < deadline )); do
+  if kubectl get --raw=/readyz --request-timeout=5s >/dev/null 2>&1; then
+    break
+  fi
+  sleep 5
+done
+
+kubectl get --raw=/readyz --request-timeout=5s >/dev/null
+kubectl wait --for=condition=Ready nodes --all --timeout=300s
+
+echo
+echo "== Estado final =="
+k3d cluster list
+kubectl get nodes
+docker ps --filter "name=${RUNNER_NAME}"
+
+echo
+echo "OK: políticas estables y runtime recuperados."

root/retire-headlamp-from-gitea.sh

@@ -0,0 +1,63 @@
+#!/usr/bin/env bash
+set -Eeuo pipefail
+IFS=$'\n\t'
+
+APPS_REPO="${APPS_REPO:-/DATA/AppData/k3d-lab/git/apps-registry}"
+PLATFORM_REPO="${PLATFORM_REPO:-/DATA/AppData/k3d-lab/git/platform-infra}"
+CONFIRM_RETIRE_HEADLAMP="${CONFIRM_RETIRE_HEADLAMP:-false}"
+
+find_application_files() {
+  local repo="$1"
+  local app_name="$2"
+
+  grep -RIl \
+    --exclude-dir=.git \
+    --include='*.yaml' \
+    --include='*.yml' \
+    -E "name:[[:space:]]*${app_name}([[:space:]]|$)" \
+    "$repo" 2>/dev/null || true
+}
+
+APP_FILES="$(find_application_files "$APPS_REPO" headlamp-app)"
+GOV_FILES="$(find_application_files "$PLATFORM_REPO" headlamp-governance)"
+
+printf 'apps-registry:\n'
+printf '  workload: %s\n' "${APPS_REPO}/workloads/headlamp"
+printf '%s\n' "$APP_FILES" | sed '/^$/d;s/^/  application: /'
+
+printf 'platform-infra:\n'
+printf '%s\n' "$GOV_FILES" | sed '/^$/d;s/^/  application: /'
+find "$PLATFORM_REPO" \
+  -path '*/.git' -prune -o \
+  -type d -name headlamp-governance -print 2>/dev/null |
+  sed 's/^/  governance: /'
+
+if [[ "$CONFIRM_RETIRE_HEADLAMP" != "true" ]]; then
+  printf '\nDRY-RUN. Usa CONFIRM_RETIRE_HEADLAMP=true para retirar las rutas detectadas.\n'
+  exit 0
+fi
+
+if [[ -d "${APPS_REPO}/workloads/headlamp" ]]; then
+  git -C "$APPS_REPO" rm -r workloads/headlamp
+fi
+
+while IFS= read -r file; do
+  [[ -n "$file" ]] || continue
+  git -C "$APPS_REPO" rm "${file#${APPS_REPO}/}"
+done <<< "$APP_FILES"
+
+while IFS= read -r file; do
+  [[ -n "$file" ]] || continue
+  git -C "$PLATFORM_REPO" rm "${file#${PLATFORM_REPO}/}"
+done <<< "$GOV_FILES"
+
+while IFS= read -r directory; do
+  [[ -n "$directory" ]] || continue
+  git -C "$PLATFORM_REPO" rm -r "${directory#${PLATFORM_REPO}/}"
+done < <(
+  find "$PLATFORM_REPO" \
+    -path '*/.git' -prune -o \
+    -type d -name headlamp-governance -print 2>/dev/null
+)
+
+printf '\nCambios preparados. Revisa git diff, haz commit y push en ambos repositorios.\n'

root/verify-stable-startup.sh

@@ -0,0 +1,62 @@
+#!/usr/bin/env bash
+set -Eeuo pipefail
+IFS=$'\n\t'
+
+CLUSTER_NAME="${CLUSTER_NAME:-lab-cluster}"
+RUNNER_NAME="${GITEA_RUNNER_CONTAINER:-gitea-runner}"
+KUBECONFIG_PATH="${KUBECONFIG_PATH:-/DATA/.kube/config}"
+
+export KUBECONFIG="$KUBECONFIG_PATH"
+
+echo "== Políticas k3d =="
+docker inspect \
+  "k3d-${CLUSTER_NAME}-server-0" \
+  "k3d-${CLUSTER_NAME}-agent-0" \
+  "k3d-${CLUSTER_NAME}-agent-1" \
+  "k3d-${CLUSTER_NAME}-serverlb" \
+  --format '{{.Name}} restart={{.HostConfig.RestartPolicy.Name}} status={{.State.Status}}'
+
+bad="$(
+  docker inspect \
+    "k3d-${CLUSTER_NAME}-server-0" \
+    "k3d-${CLUSTER_NAME}-agent-0" \
+    "k3d-${CLUSTER_NAME}-agent-1" \
+    "k3d-${CLUSTER_NAME}-serverlb" \
+    --format '{{.HostConfig.RestartPolicy.Name}}' |
+  grep -v '^unless-stopped$' || true
+)"
+[[ -z "$bad" ]] || { echo "ERROR: k3d debe usar unless-stopped." >&2; exit 1; }
+
+echo
+echo "== Runner =="
+docker inspect "$RUNNER_NAME" \
+  --format '{{.Name}} restart={{.HostConfig.RestartPolicy.Name}} status={{.State.Status}}'
+
+[[ "$(docker inspect -f '{{.HostConfig.RestartPolicy.Name}}' "$RUNNER_NAME")" == "unless-stopped" ]]
+[[ "$(docker inspect -f '{{.State.Running}}' "$RUNNER_NAME")" == "true" ]]
+
+echo
+echo "== Docker independiente =="
+test ! -f /etc/systemd/system/docker.service.d/10-k3d-storage.conf
+echo "OK"
+
+echo
+echo "== Timer =="
+systemctl is-enabled k3d-lab-ensure.timer
+systemctl is-active k3d-lab-ensure.timer
+systemctl cat k3d-lab-ensure.timer | grep -E 'OnBootSec=5min|Persistent=false'
+
+echo
+echo "== Kubernetes =="
+k3d cluster list
+kubectl get nodes
+
+ready="$(
+  kubectl get nodes \
+    -o jsonpath='{range .items[*]}{.status.conditions[?(@.type=="Ready")].status}{"\n"}{end}' |
+  grep -c '^True$' || true
+)"
+[[ "$ready" -eq 3 ]] || { echo "ERROR: nodos Ready=${ready}, esperado=3." >&2; exit 1; }
+
+echo
+echo "OK: listo para reinicio con orden estable."