[Feature] [PD Disaggregation] simplify configuration for pd-disaggregated deployment, and refactor post-init and usage for all ports (#5415)

* [feat] simplify configuration for pd-disaggregated deployment, and refactor post-init and usage for all ports

* [fix] fix some bugs

* [fix] fix rdma port for cache manager/messager

* [fix] temporarily cancel port availability check to see if it can pass ci test

* [feat] simplify args for multi api server

* [fix] fix dp

* [fix] fix port for xpu

* [fix] add tests for ports post processing & fix ci

* [test] fix test_multi_api_server

* [fix] fix rdma_comm_ports args for multi_api_server

* [fix] fix test_common_engine

* [fix] fix test_cache_transfer_manager

* [chore] automatically setting FD_ENABLE_MULTI_API_SERVER

* [fix] avoid api server from creating engine_args twice

* [fix] fix test_run_batch

* [fix] fix test_metrics

* [fix] fix splitwise connector init

* [test] add test_rdma_transfer and test_expert_service

* [fix] fix code syntax

* [fix] fix test_rdma_transfer and build wheel with rdma script
This commit is contained in:
Yonghua Li
2025-12-17 15:50:42 +08:00
committed by GitHub
parent cdc0004894
commit 0c8c6369ed
34 changed files with 1323 additions and 409 deletions
@@ -85,7 +85,7 @@ def parse_args():
help="engine worker queue port",
)
parser.add_argument("--num_cpu_blocks", type=int, default=4, help="cpu cache block number")
parser.add_argument("--engine_pid", type=str, default=None, help="engine pid")
parser.add_argument("--ipc_suffix", type=str, default=None, help="engine pid")
parser.add_argument(
"--protocol",
type=str,
@@ -140,7 +140,7 @@ class CacheTransferManager:
self.n_ranks = args.mp_num
self.rank = rank
self.device = device
self.engine_pid = args.engine_pid
self.ipc_suffix = args.ipc_suffix
self.cache_dtype = args.cache_dtype
address = (args.pod_ip, args.cache_queue_port)
@@ -157,7 +157,7 @@ class CacheTransferManager:
name="cache_ready_signal",
array=cache_ready_signal_data,
dtype=np.int32,
suffix=self.engine_pid,
suffix=self.ipc_suffix,
create=False,
)
swap_space_ready_data = np.zeros(shape=[args.mp_num], dtype=np.int32)
@@ -165,7 +165,7 @@ class CacheTransferManager:
name="swap_space_ready_signal",
array=swap_space_ready_data,
dtype=np.int32,
suffix=self.engine_pid,
suffix=self.ipc_suffix,
create=False,
)
@@ -180,7 +180,7 @@ class CacheTransferManager:
name="cache_task_broadcast_signal",
array=cache_task_broadcast_data,
dtype=np.int32,
suffix=args.engine_pid,
suffix=args.ipc_suffix,
create=False,
)
@@ -653,7 +653,7 @@ class CacheTransferManager:
name="kv_cache_status",
array=kv_cache_status,
dtype=np.int32,
suffix=self.engine_pid,
suffix=self.ipc_suffix,
create=False,
)
while True: