[RL] Add --nccl-port to prevent port conflict (#7418)

python/sglang/srt/server_args.py

@@ -68,6 +68,7 @@ class ServerArgs:
     # Port for the HTTP server
     host: str = "127.0.0.1"
     port: int = 30000
+    nccl_port: Optional[int] = None
 
     # Memory and scheduling
     mem_fraction_static: Optional[float] = None
@@ -599,6 +600,12 @@ class ServerArgs:
             default=ServerArgs.port,
             help="The port of the HTTP server.",
         )
+        parser.add_argument(
+            "--nccl-port",
+            type=int,
+            default=ServerArgs.nccl_port,
+            help="The port for NCCL distributed environment setup. Defaults to a random port.",
+        )
         parser.add_argument(
             "--tokenizer-mode",
             type=str,
@@ -1747,6 +1754,7 @@ class PortArgs:
 
     @staticmethod
     def init_new(server_args, dp_rank: Optional[int] = None) -> "PortArgs":
+        if server_args.nccl_port is None:
             port = server_args.port + random.randint(100, 1000)
             while True:
                 if is_port_available(port):
@@ -1755,6 +1763,8 @@ class PortArgs:
                     port += 42
                 else:
                     port -= 43
+        else:
+            port = server_args.nccl_port
 
         if not server_args.enable_dp_attention:
             # Normal case, use IPC within a single node