Make engine core client handshake timeout configurable (#27444)
Signed-off-by: Seiji Eicher <seiji@anyscale.com>
This commit is contained in:
@@ -20,6 +20,7 @@ import zmq
|
||||
import zmq.asyncio
|
||||
|
||||
from vllm.config import VllmConfig
|
||||
from vllm.envs import VLLM_ENGINE_READY_TIMEOUT_S
|
||||
from vllm.logger import init_logger
|
||||
from vllm.lora.request import LoRARequest
|
||||
from vllm.tasks import SupportedTask
|
||||
@@ -528,7 +529,9 @@ class MPClient(EngineCoreClient):
|
||||
identities = set(self.core_engines)
|
||||
sync_input_socket = zmq.Socket.shadow(self.input_socket)
|
||||
while identities:
|
||||
if not sync_input_socket.poll(timeout=600_000):
|
||||
if not sync_input_socket.poll(
|
||||
timeout=VLLM_ENGINE_READY_TIMEOUT_S * 1000 # convert to ms
|
||||
):
|
||||
raise TimeoutError(
|
||||
"Timed out waiting for engines to send"
|
||||
"initial message on input socket."
|
||||
@@ -1340,7 +1343,9 @@ class DPLBAsyncMPClient(DPAsyncMPClient):
|
||||
# Wait for ready messages from new engines on the input socket
|
||||
sync_input_socket = zmq.Socket.shadow(self.input_socket)
|
||||
while new_engine_identities:
|
||||
if not sync_input_socket.poll(timeout=600_000):
|
||||
if not sync_input_socket.poll(
|
||||
timeout=VLLM_ENGINE_READY_TIMEOUT_S * 1000 # convert to ms
|
||||
):
|
||||
raise TimeoutError(
|
||||
"Timed out waiting for new engines to send initial "
|
||||
"message on input socket."
|
||||
|
||||
Reference in New Issue
Block a user