[PD] Fix failure abort (#6535)

python/sglang/srt/managers/schedule_batch.py

@@ -48,7 +48,9 @@ from sglang.global_config import global_config
 from sglang.srt.configs.model_config import ModelConfig
 from sglang.srt.constrained.base_grammar_backend import BaseGrammarObject
 from sglang.srt.disaggregation.base import BaseKVSender
-from sglang.srt.disaggregation.decode import ScheduleBatchDisaggregationDecodeMixin
+from sglang.srt.disaggregation.decode_schedule_batch_mixin import (
+    ScheduleBatchDisaggregationDecodeMixin,
+)
 from sglang.srt.layers.multimodal import gpu_tensor_hash
 from sglang.srt.mem_cache.base_prefix_cache import BasePrefixCache
 from sglang.srt.mem_cache.chunk_cache import ChunkCache

python/sglang/srt/managers/scheduler.py

@@ -582,6 +582,8 @@ class Scheduler(
                 gloo_group=self.attn_tp_cpu_group,
                 req_to_metadata_buffer_idx_allocator=req_to_metadata_buffer_idx_allocator,
                 metadata_buffers=metadata_buffers,
+                scheduler=self,
+                tree_cache=self.tree_cache,
             )
 
             # The decode requests pending for pre-allocation