diff --git a/vllm_ascend/distributed/kv_transfer/kv_p2p/mooncake_layerwise_connector.py b/vllm_ascend/distributed/kv_transfer/kv_p2p/mooncake_layerwise_connector.py index e06881fe..ee161c94 100644 --- a/vllm_ascend/distributed/kv_transfer/kv_p2p/mooncake_layerwise_connector.py +++ b/vllm_ascend/distributed/kv_transfer/kv_p2p/mooncake_layerwise_connector.py @@ -567,10 +567,26 @@ class MooncakeLayerwiseConnectorScheduler: self._reqs_need_recv: dict[str, tuple[Request, list[int], list[int]]] = {} self._reqs_need_send_layerwise: dict[str, SendReqInfo] = {} - self.executor = ThreadPoolExecutor(32) - self.metaserver_client = httpx.Client( - limits=httpx.Limits(max_connections=100000), timeout=None) + tls_config: dict[ + str, Any] = vllm_config.kv_transfer_config.get_from_extra_config( + "tls_config", {}) + ssl_keyfile = tls_config.get("ssl_keyfile", None) + ssl_certfile = tls_config.get("ssl_certfile", None) + ssl_ca_certs = tls_config.get("ssl_ca_certs", False) + ssl_keyfile_password = tls_config.get("ssl_keyfile_password", None) + self.cert_path = (ssl_certfile, ssl_keyfile, ssl_keyfile_password) + self.ssl_enable = tls_config.get("ssl_enable", False) + self.ca_path = ssl_ca_certs + if self.ssl_enable: + self.metaserver_client = httpx.Client( + limits=httpx.Limits(max_connections=100000), + timeout=None, + cert=self.cert_path, + verify=self.ca_path) + else: + self.metaserver_client = httpx.Client( + limits=httpx.Limits(max_connections=100000), timeout=None) def get_num_new_matched_tokens( self, request: "Request", @@ -645,11 +661,10 @@ class MooncakeLayerwiseConnectorScheduler: remote_host=self.side_channel_host, remote_port=self.side_channel_port, ) - future = self.executor.submit( - self._access_metaserver, - url=params.get("metaserver", None), - message=kv_transfer_params, - ) + + future = self.executor.submit(self._access_metaserver, + url=params.get("metaserver", None), + message=kv_transfer_params) def handle_exception(future): if future.exception():