fix(aiops): escalate failed host verification

apps/api/src/services/auto_repair_service.py

@@ -521,6 +521,13 @@ class AutoRepairService:
                         # 2026-04-27 P3.1-T1 by Claude — 三 Tier-1 服務整合
                         # PostExecutionVerifier 判斷失敗/降級 → 觸發自動 Rollback
                         if verification_result in ("failed", "degraded"):
+                            if self._should_escalate_failed_verification(incident, playbook):
+                                await self._escalate_failed_verification(
+                                    incident=incident,
+                                    playbook=playbook,
+                                    verification_result=verification_result,
+                                )
+                                return
                             try:
                                 from src.services.rollback_manager import get_rollback_manager
                                 from src.services.declarative_remediation import DeclarativeRemediation
@@ -716,6 +723,69 @@ class AutoRepairService:
                 return True
         return False
 
+    def _should_escalate_failed_verification(self, incident: Incident, playbook: Playbook) -> bool:
+        """非 K8s 修復或主機/備份事件驗證失敗時，禁止合成 K8s rollback。"""
+
+        return self._is_host_or_backup_incident(incident) or not self._playbook_has_k8s_steps(playbook)
+
+    async def _escalate_failed_verification(
+        self,
+        *,
+        incident: Incident,
+        playbook: Playbook,
+        verification_result: str,
+    ) -> None:
+        """Post-verification failed but rollback is unsafe; notify emergency channel."""
+
+        target = (incident.affected_services or ["unknown"])[0]
+        namespace = "awoooi-prod"
+        alert_type = self._incident_alert_type(incident)
+        reason = (
+            f"auto repair playbook {playbook.playbook_id} verification={verification_result}; "
+            "rollback is unsafe for host/backup or non-K8s remediation"
+        )
+        logger.warning(
+            "auto_repair_verification_failed_emergency",
+            incident_id=incident.incident_id,
+            playbook_id=playbook.playbook_id,
+            verification_result=verification_result,
+            target=target,
+        )
+        try:
+            from src.services.emergency_escalation_service import (
+                escalate_auto_repair_unavailable,
+            )
+
+            await escalate_auto_repair_unavailable(
+                incident_id=incident.incident_id,
+                approval_id=None,
+                alert_type=alert_type,
+                target_resource=target,
+                namespace=namespace,
+                failure_reason=reason,
+                attempted_actions=(
+                    f"auto_repair:{playbook.playbook_id} -> verifier:{verification_result} "
+                    "-> emergency_intervention"
+                ),
+            )
+        except Exception as exc:
+            logger.warning(
+                "auto_repair_verification_emergency_failed",
+                incident_id=incident.incident_id,
+                playbook_id=playbook.playbook_id,
+                error=str(exc),
+            )
+
+    def _incident_alert_type(self, incident: Incident) -> str:
+        """Best-effort alertname for emergency cards."""
+
+        for signal in incident.signals or []:
+            labels = signal.labels or {}
+            alertname = labels.get("alertname") or signal.alert_name
+            if alertname:
+                return str(alertname)
+        return "AutoRepairVerificationFailed"
+
     def _risk_exceeds_threshold(self, risk: RiskLevel) -> bool:
         """檢查風險是否超過自動修復門檻"""
         high_risks = {RiskLevel.HIGH, RiskLevel.CRITICAL}

apps/api/src/services/emergency_escalation_service.py

@@ -19,7 +19,7 @@ logger = structlog.get_logger(__name__)
 async def escalate_auto_repair_unavailable(
     *,
     incident_id: str,
-    approval_id: str,
+    approval_id: str | None,
     alert_type: str,
     target_resource: str,
     namespace: str,