deploy fix bug

ops-scripts/tencent_operator.py

@@ -52,6 +52,12 @@ def sdk_body(resp: Any) -> dict[str, Any]:
     return body
 
 
+def should_retry_without_output_cos(exc: TencentCloudSDKException) -> bool:
+    code = str(getattr(exc, "code", "") or "")
+    message = str(getattr(exc, "message", "") or "")
+    return code == "ResourceNotFound.RoleNotFound" and "TAT_QCSLinkedRoleInUploadInvocation" in message
+
+
 class CloudOperator:
     def __init__(self, config: dict[str, Any], service_name: str, instance_id: str, release_id: str | None) -> None:
         self.config = config
@@ -201,9 +207,25 @@ class CloudOperator:
             payload["OutputCOSBucketUrl"] = self.cos_bucket_url()
             payload["OutputCOSKeyPrefix"] = f"{output_prefix}/{self.release_id or 'adhoc'}/{self.service_name}/{self.instance_id}"
 
+        return self._run_tat_command(payload, allow_retry_without_output_cos=bool(output_prefix))
+
+    def _run_tat_command(self, payload: dict[str, Any], allow_retry_without_output_cos: bool) -> str:
         req = tat_models.RunCommandRequest()
         req.from_json_string(json.dumps(payload))
+        try:
             resp = self.tat.RunCommand(req)
+        except TencentCloudSDKException as exc:
+            if allow_retry_without_output_cos and should_retry_without_output_cos(exc):
+                log(
+                    "TAT missing linked role TAT_QCSLinkedRoleInUploadInvocation, "
+                    "retrying RunCommand without OutputCOSBucketUrl"
+                )
+                retry_payload = dict(payload)
+                retry_payload.pop("OutputCOSBucketUrl", None)
+                retry_payload.pop("OutputCOSKeyPrefix", None)
+                return self._run_tat_command(retry_payload, allow_retry_without_output_cos=False)
+            raise
+
         body = sdk_body(resp)
         return str(body["InvocationId"])