From 3397f5a9aa6650e57b57f39e159f3a91bbefe797 Mon Sep 17 00:00:00 2001
From: Your Name <you@example.com>
Date: Tue, 30 Jun 2026 21:26:47 +0800
Subject: [PATCH] fix(recovery): open controlled drain lane after guardrails

---
 scripts/reboot-recovery/awoooi-startup-110.sh | 32 ++++++++++++++++++-
 .../test_cold_start_monitor_bounded_probes.py | 17 ++++++++++
 2 files changed, 48 insertions(+), 1 deletion(-)

diff --git a/scripts/reboot-recovery/awoooi-startup-110.sh b/scripts/reboot-recovery/awoooi-startup-110.sh
index bc6a9a78..01c97f1d 100644
--- a/scripts/reboot-recovery/awoooi-startup-110.sh
+++ b/scripts/reboot-recovery/awoooi-startup-110.sh
@@ -509,6 +509,13 @@ cd_lane_drain_is_controlled_available() {
     return 0
 }
 
+cd_lane_root_restore_sources_left() {
+    find /root -maxdepth 1 -type d \( \
+        -name 'awoooi-cd-lane-disabled-*' -o \
+        -name 'awoooi-cd-lane-drain-disabled-*' \
+    \) -print 2>/dev/null | wc -l | tr -d " "
+}
+
 quarantine_cd_lane_registration_fail_closed() {
     local quarantine_dir
     local lane_dir
@@ -625,10 +632,25 @@ ensure_host_runner_fail_closed() {
     pkill -KILL -f "^${RUNNER_DIR}/act_runner(\\.real-[^ ]*)? daemon" >/dev/null 2>&1 || true
     quarantine_cd_lane_root_restore_sources_fail_closed
     for binary in "${RUNNER_FAIL_CLOSED_BINARY_PATHS[@]}"; do
+        if [ "$START_CD_LANE_ALLOWED" = "1" ] && [ "$binary" = "$CD_LANE_DRAIN_BINARY" ]; then
+            continue
+        fi
         guard_runner_binary_fail_closed "$binary"
     done
 }
 
+CD_LANE_ROOT_RESTORE_LEFT="$(cd_lane_root_restore_sources_left)"
+if [ "$START_CONTROLLED_CD_LANE" = "1" ]; then
+    if ! cd_lane_drain_is_controlled_available; then
+        log "⛔ AWOOOI_START_CONTROLLED_CD_LANE=1 但 controlled drain lane config/binary 未通過；維持 fail-closed"
+    elif [ "$CD_LANE_ROOT_RESTORE_LEFT" != "0" ]; then
+        log "⛔ AWOOOI_START_CONTROLLED_CD_LANE=1 但 root restore-source left=${CD_LANE_ROOT_RESTORE_LEFT}；維持 fail-closed"
+    else
+        START_CD_LANE_ALLOWED=1
+        log "✅ controlled cd-lane drain preflight passed; legacy runner 仍維持 fail-closed"
+    fi
+fi
+
 if [ -x "$RUNNER_DIR/act_runner" ] && [ -f "$RUNNER_DIR/config.yaml" ]; then
     # 若舊的 .runner 配置指向過期 hostname，只有在明確允許啟動 runner
     # 時才清除重新註冊；預設降壓模式不得碰 registration 狀態。
@@ -729,7 +751,15 @@ else
     log "⚠️  找不到 act-runner binary/config: $RUNNER_DIR"
 fi
 
-log "✅ controlled cd-lane startup override active; startup will not enforce drain fail-closed"
+if [ "$START_CD_LANE_ALLOWED" = "1" ]; then
+    install_controlled_cd_lane_drain_unit
+    systemctl daemon-reload >/dev/null 2>&1 || true
+    systemctl unmask "$CD_LANE_DRAIN_SERVICE" >/dev/null 2>&1 || true
+    systemctl enable --now "$CD_LANE_DRAIN_SERVICE" >/dev/null 2>&1 || true
+    ensure_controlled_cd_lane_open
+else
+    log "✅ controlled cd-lane remains closed unless AWOOOI_START_CONTROLLED_CD_LANE=1 passes guardrails"
+fi
 
 # ──────────────────────────────────────────────
 # STEP 7: Sentry（Error Tracking）
diff --git a/scripts/reboot-recovery/tests/test_cold_start_monitor_bounded_probes.py b/scripts/reboot-recovery/tests/test_cold_start_monitor_bounded_probes.py
index e09c2286..0c44b1eb 100644
--- a/scripts/reboot-recovery/tests/test_cold_start_monitor_bounded_probes.py
+++ b/scripts/reboot-recovery/tests/test_cold_start_monitor_bounded_probes.py
@@ -76,6 +76,23 @@ def test_startup_110_quarantines_corrupt_docker_container_metadata() -> None:
     assert 'run_bounded "$DOCKER_START_TIMEOUT_SECONDS" systemctl start docker.socket docker.service' in text
 
 
+def test_startup_110_opens_only_controlled_cd_lane_after_guardrails() -> None:
+    text = STARTUP_110.read_text(encoding="utf-8")
+
+    assert 'START_CONTROLLED_CD_LANE="${AWOOOI_START_CONTROLLED_CD_LANE:-0}"' in text
+    assert "cd_lane_root_restore_sources_left()" in text
+    assert 'CD_LANE_ROOT_RESTORE_LEFT="$(cd_lane_root_restore_sources_left)"' in text
+    assert 'START_CD_LANE_ALLOWED=1' in text
+    assert 'install_controlled_cd_lane_drain_unit' in text
+    assert 'systemctl unmask "$CD_LANE_DRAIN_SERVICE"' in text
+    assert 'systemctl enable --now "$CD_LANE_DRAIN_SERVICE"' in text
+    assert 'ensure_controlled_cd_lane_open' in text
+    assert 'if [ "$START_CD_LANE_ALLOWED" = "1" ] && [ "$binary" = "$CD_LANE_DRAIN_BINARY" ]; then' in text
+    assert 'systemctl enable --now "$RUNNER_SERVICE"' in text
+    assert "legacy runner 仍維持 fail-closed" in text
+    assert "controlled cd-lane remains closed unless AWOOOI_START_CONTROLLED_CD_LANE=1 passes guardrails" in text
+
+
 def test_cold_start_deploy_parity_verifier_bounds_ssh_readback() -> None:
     text = VERIFY_DEPLOY.read_text(encoding="utf-8")