[Bugfix] fix ZeroDivisionError when prefill_tp_size > num_kv_head and fix tp_resharding README (#3437)

### What this PR does / why we need it?
Fix ZeroDivisionError when prefill_tp_size > num_kv_head, in this
situation, num_head_replica can be 0 and used to divide another value,
this PR restricts the minimum value of a to be 1. And this PR fix
tp_resharding README.

### Does this PR introduce _any_ user-facing change?
No.

### How was this patch tested?
By CI.

- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

---------

Signed-off-by: liziyu <liziyu16@huawei.com>
Signed-off-by: nwpu-zxr <zhouxuerong2@huawei.com>
Co-authored-by: liziyu <liziyu16@huawei.com>

docs/source/tutorials/multi_node_pd_disaggregation_mooncake.md

@@ -114,10 +114,10 @@ export VLLM_USE_V1=1
 export HCCL_BUFFSIZE=1024
 export OMP_PROC_BIND=false
 export OMP_NUM_THREADS=10
-export ASCEND_AGGREGATE_ENABLE=1
-export ASCEND_TRANSPORT_PRINT=0
-export ACL_OP_INIT_MODE=1
-export ASCEND_A3_ENABLE=1
+export ASCEND_AGGREGATE_ENABLE=1  # enable aggregated transmission
+export ASCEND_TRANSPORT_PRINT=0  # print ascend transport logs
+export ACL_OP_INIT_MODE=1  # acl op initialization mode to prevent device id acquisition failure
+export ASCEND_A3_ENABLE=1  # enable hccs transmission for A3; set to 0 for A2
 export LD_LIBRARY_PATH=/usr/local/Ascend/ascend-toolkit/latest/python/site-packages:$LD_LIBRARY_PATH
 
 vllm serve /model/Qwen3-235B-A22B-W8A8 \
@@ -137,7 +137,6 @@ vllm serve /model/Qwen3-235B-A22B-W8A8 \
   --max-model-len 32768 \
   --max-num-batched-tokens 32768 \
   --trust-remote-code \
-  --no-enable-prefix-caching \
   --gpu-memory-utilization 0.9 \
   --kv-transfer-config \
   '{"kv_connector": "MooncakeLayerwiseConnector",
@@ -197,7 +196,6 @@ vllm serve /model/Qwen3-235B-A22B-W8A8 \
   --max-model-len 32768 \
   --max-num-batched-tokens 32768 \
   --trust-remote-code \
-  --no-enable-prefix-caching \
   --gpu-memory-utilization 0.9 \
   --kv-transfer-config \
   '{"kv_connector": "MooncakeLayerwiseConnector",
@@ -363,6 +361,10 @@ export VLLM_USE_V1=1
 export HCCL_BUFFSIZE=1024
 export OMP_PROC_BIND=false
 export OMP_NUM_THREADS=10
+export ASCEND_AGGREGATE_ENABLE=1
+export ASCEND_TRANSPORT_PRINT=0
+export ACL_OP_INIT_MODE=1
+export ASCEND_A3_ENABLE=1
 export LD_LIBRARY_PATH=/usr/local/Ascend/ascend-toolkit/latest/python/site-packages:$LD_LIBRARY_PATH
 
 vllm serve /model/Qwen3-235B-A22B-W8A8 \
@@ -382,7 +384,6 @@ vllm serve /model/Qwen3-235B-A22B-W8A8 \
   --max-model-len 32768 \
   --max-num-batched-tokens 32768 \
   --trust-remote-code \
-  --no-enable-prefix-caching \
   --gpu-memory-utilization 0.9 \
   --kv-transfer-config \
   '{"kv_connector": "MooncakeConnector",
@@ -419,6 +420,10 @@ export VLLM_USE_V1=1
 export HCCL_BUFFSIZE=1024
 export OMP_PROC_BIND=false
 export OMP_NUM_THREADS=10
+export ASCEND_AGGREGATE_ENABLE=1
+export ASCEND_TRANSPORT_PRINT=0
+export ACL_OP_INIT_MODE=1
+export ASCEND_A3_ENABLE=1
 export LD_LIBRARY_PATH=/usr/local/Ascend/ascend-toolkit/latest/python/site-packages:$LD_LIBRARY_PATH
 
 vllm serve /model/Qwen3-235B-A22B-W8A8 \
@@ -438,7 +443,6 @@ vllm serve /model/Qwen3-235B-A22B-W8A8 \
   --max-model-len 32768 \
   --max-num-batched-tokens 32768 \
   --trust-remote-code \
-  --no-enable-prefix-caching \
   --gpu-memory-utilization 0.9 \
   --kv-transfer-config \
   '{"kv_connector": "MooncakeConnector",
@@ -475,6 +479,10 @@ export VLLM_USE_V1=1
 export HCCL_BUFFSIZE=2048
 export OMP_PROC_BIND=false
 export OMP_NUM_THREADS=10
+export ASCEND_AGGREGATE_ENABLE=1
+export ASCEND_TRANSPORT_PRINT=0
+export ACL_OP_INIT_MODE=1
+export ASCEND_A3_ENABLE=1
 export LD_LIBRARY_PATH=/usr/local/Ascend/ascend-toolkit/latest/python/site-packages:$LD_LIBRARY_PATH
 
 vllm serve /model/Qwen3-235B-A22B-W8A8 \
@@ -532,6 +540,10 @@ export VLLM_USE_V1=1
 export HCCL_BUFFSIZE=2048
 export OMP_PROC_BIND=false
 export OMP_NUM_THREADS=10
+export ASCEND_AGGREGATE_ENABLE=1
+export ASCEND_TRANSPORT_PRINT=0
+export ACL_OP_INIT_MODE=1
+export ASCEND_A3_ENABLE=1
 export LD_LIBRARY_PATH=/usr/local/Ascend/ascend-toolkit/latest/python/site-packages:$LD_LIBRARY_PATH
 
 vllm serve /model/Qwen3-235B-A22B-W8A8 \

tests/ut/kv_connector/test_mooncake_layerwise_connector.py

@@ -79,7 +79,7 @@ class TestKVCacheSendingLayerThreadBasic(unittest.TestCase):
         self.p1 = patch(
             'vllm_ascend.distributed.mooncake_layerwise_connector.get_ascend_config',
             new=MagicMock(return_value=SimpleNamespace(
-                pd_tp_ratio=1, num_head_replica=0, pd_head_ratio=1)))
+                pd_tp_ratio=1, num_head_replica=1, pd_head_ratio=1)))
         self.p2 = patch(
             'vllm_ascend.distributed.mooncake_layerwise_connector.get_current_vllm_config',
             new=MagicMock(return_value=SimpleNamespace(
@@ -244,7 +244,7 @@ class TestSendingLayerThread(unittest.TestCase):
         self.p1 = patch(
             'vllm_ascend.distributed.mooncake_layerwise_connector.get_ascend_config',
             new=MagicMock(return_value=SimpleNamespace(
-                pd_tp_ratio=1, num_head_replica=0, pd_head_ratio=1)))
+                pd_tp_ratio=1, num_head_replica=1, pd_head_ratio=1)))
         self.p2 = patch(
             'vllm_ascend.distributed.mooncake_layerwise_connector.get_current_vllm_config',
             new=MagicMock(return_value=SimpleNamespace(
@@ -903,7 +903,7 @@ class TestMooncakeLayerwiseConnectorWorker(unittest.TestCase):
             patch(
                 'vllm_ascend.distributed.mooncake_layerwise_connector.get_ascend_config',
                 return_value=SimpleNamespace(pd_tp_ratio=1,
-                                             num_head_replica=0,
+                                             num_head_replica=1,
                                              pd_head_ratio=1),
             ),
             patch(

vllm_ascend/ascend_config.py

@@ -102,7 +102,7 @@ class AscendConfig:
                 )
         self.pd_tp_ratio = 1
         self.pd_head_ratio = 1
-        self.num_head_replica = 0
+        self.num_head_replica = 1
         if vllm_config.kv_transfer_config is not None and not vllm_config.model_config.is_deepseek_mla:
             prefill_tp_size = vllm_config.kv_transfer_config.get_from_extra_config(
                 "prefill", {"tp_size": 1})["tp_size"]
@@ -115,7 +115,7 @@ class AscendConfig:
                     # only support Qwen model now
                     # TODO: use a more robust method to get kv_head_num
                     num_kv_head = vllm_config.model_config.hf_config.num_key_value_heads
-                    self.num_head_replica = prefill_tp_size // num_kv_head
+                    self.num_head_replica = prefill_tp_size // num_kv_head if prefill_tp_size >= num_kv_head else 1
                     prefill_tp_size = min(prefill_tp_size, num_kv_head)
                     decode_tp_size = min(decode_tp_size, num_kv_head)
                     self.pd_head_ratio = prefill_tp_size // decode_tp_size

vllm_ascend/distributed/mooncake_layerwise_connector.py

@@ -360,7 +360,7 @@ class SendingLayerThread(threading.Thread):
         remote_kv_base_addrs = req_meta.kv_caches_base_addr
 
         remote_block_ids = req_meta.block_ids
-        if self.num_head_replica >= 1 and self.tp_rank % self.num_head_replica != 0:
+        if self.tp_rank % self.num_head_replica != 0:
             pass
         elif self.pd_head_ratio == 1:
             layer_local_kv_base_addr = [