Remove temporary CP cache-hit trace instrumentation

The GSM8K/cache-hit debugging pass added request-correlation logs across scheduler, radix, HiCache, Mooncake, and prefill handoff. The root cause work has moved to request-slot remap semantics, so those high-cardinality traces are no longer needed in the runtime diff.

This removes the temporary tracing helpers and call sites while leaving existing fail-fast checks, fallback warnings, and explicit debug/timing infrastructure that is still part of normal CP shared-KV diagnostics.

Constraint: Production CP hot paths should not carry investigation-only request signatures or transfer summaries
Rejected: Keep all debug logs gated by env | even gated logs increase maintenance surface and encourage stale diagnosis paths
Confidence: high
Scope-risk: moderate
Directive: Reintroduce request-correlation logs only as a narrow opt-in probe with a planned removal point
Tested: Local py_compile for touched runtime files; remote py_compile; remote pytest test_cp_shared_kv_runtime.py, test_nsa_cp_utils.py, test_cp_shared_kv_layout.py => 263 passed, 5 warnings, 2 subtests passed
Not-tested: Long-running ETE log-volume comparison after removal

python/sglang/srt/disaggregation/mooncake/conn.py

@@ -64,21 +64,6 @@ def _cp_draft_shared_kv_debug(message: str, *args, limit: int = 64) -> None:
     logger.info("[CP_DRAFT_SHARED_KV] " + message, *args)
 
 
-def _cp_shared_kv_bs_gt1_mooncake_debug(
-    key: str, message: str, *args, limit: Optional[int] = None
-) -> None:
-    if not envs.SGLANG_CP_SHARED_KV_BS_GT1_DEBUG.get():
-        return
-    if limit is None:
-        limit = int(envs.SGLANG_CP_SHARED_KV_BS_GT1_DEBUG_LIMIT.get())
-    count_key = "bs_gt1:" + key
-    count = _CP_SHARED_DEBUG_COUNTS.get(count_key, 0)
-    if limit > 0 and count >= limit:
-        return
-    _CP_SHARED_DEBUG_COUNTS[count_key] = count + 1
-    logger.info("[CP_SHARED_KV_BS_GT1_DEBUG] event=%s " + message, key, *args)
-
-
 def _mooncake_transfer_stats_enabled() -> bool:
     return envs.SGLANG_DISAGGREGATION_TRANSFER_STATS.get()
 
@@ -1226,19 +1211,6 @@ class MooncakeKVManager(CommonKVManager):
                             _np_summary(chunked_dst_kv_indice),
                             kv_chunk.is_last_chunk,
                         )
-                        _cp_shared_kv_bs_gt1_mooncake_debug(
-                            "transfer_worker_kv",
-                            "cp_rank=%s room=%s prefill_pages=%s "
-                            "logical_positions=%s dst_pages=%s is_last=%s "
-                            "session=%s",
-                            self.attn_cp_rank,
-                            kv_chunk.room,
-                            _np_summary(kv_chunk.prefill_kv_indices),
-                            _np_summary(kv_chunk.logical_page_positions),
-                            _np_summary(chunked_dst_kv_indice),
-                            kv_chunk.is_last_chunk,
-                            req.mooncake_session_id,
-                        )
                         if envs.SGLANG_DEBUG_CP_SHARED_KV.get():
                             _cp_shared_debug_log(
                                 "transfer_worker_kv",
@@ -1354,18 +1326,6 @@ class MooncakeKVManager(CommonKVManager):
                                         ),
                                         _np_summary(dst_state_indices),
                                     )
-                                _cp_shared_kv_bs_gt1_mooncake_debug(
-                                    "transfer_worker_state",
-                                    "cp_rank=%s room=%s prefill_state_pages=%s "
-                                    "state_positions=%s dst_state_pages=%s "
-                                    "session=%s",
-                                    self.attn_cp_rank,
-                                    kv_chunk.room,
-                                    _np_summary(kv_chunk.state_indices),
-                                    _np_summary(kv_chunk.state_logical_page_positions),
-                                    _np_summary(dst_state_indices),
-                                    req.mooncake_session_id,
-                                )
                                 self.maybe_send_extra(
                                     req,
                                     kv_chunk.state_indices,
@@ -1746,22 +1706,6 @@ class MooncakeKVSender(CommonKVSender):
                 is_last_chunk,
                 getattr(self.kv_mgr.kv_args, "draft_kv_buffer_count", None),
             )
-            _cp_shared_kv_bs_gt1_mooncake_debug(
-                "sender_filter",
-                "cp_rank=%s room=%s page_start=%s orig_kv_pages=%s "
-                "filtered_kv_pages=%s kv_positions=%s orig_state_pages=%s "
-                "filtered_state_pages=%s state_positions=%s is_last=%s",
-                self.kv_mgr.attn_cp_rank,
-                self.bootstrap_room,
-                chunk_page_start,
-                _np_summary(orig_kv_indices),
-                _np_summary(kv_indices),
-                _np_summary(logical_page_positions),
-                _np_summary(orig_state_indices),
-                _np_summary(state_indices),
-                _np_summary(state_logical_page_positions),
-                is_last_chunk,
-            )
         # Special handling for cp
         elif self.kv_mgr.enable_all_cp_ranks_for_transfer:
             kv_indices, index_slice = filter_kv_indices_for_cp_rank(

python/sglang/srt/disaggregation/prefill.py

@@ -43,7 +43,6 @@ from sglang.srt.disaggregation.utils import (
     kv_to_page_num,
     poll_and_all_reduce_attn_cp_tp_group,
     prepare_abort,
-    token_ids_debug_signature,
 )
 from sglang.srt.managers.schedule_batch import (
     FINISH_ABORT,
@@ -1305,7 +1304,7 @@ class SchedulerDisaggregationPrefillMixin:
                 "rid=%s room=%s start_idx=%s end_idx=%s last_chunk=%s "
                 "page_size=%s pages=%s state_pages=%s prefix_len=%s "
                 "host_hit_length=%s extend_input_len=%s fill_len=%s "
-                "origin_input_len=%s has_draft_pool=%s origin_sig=%s fill_sig=%s",
+                "origin_input_len=%s has_draft_pool=%s",
                 req.rid,
                 req.bootstrap_room,
                 start_idx,
@@ -1320,8 +1319,6 @@ class SchedulerDisaggregationPrefillMixin:
                 len(req.fill_ids),
                 len(req.origin_input_ids),
                 has_draft_pool,
-                token_ids_debug_signature(getattr(req, "origin_input_ids", None)),
-                token_ids_debug_signature(getattr(req, "fill_ids", None)),
             )
         if has_draft_pool and draft_prefix_overlap > 0:
             _cp_draft_shared_kv_debug(

python/sglang/srt/disaggregation/utils.py

@@ -67,31 +67,6 @@ def _cp_shared_kv_poll_debug_enabled() -> bool:
     return envs.SGLANG_CP_SHARED_KV_BS_GT1_DEBUG.get()
 
 
-def token_ids_debug_signature(token_ids: Any, *, sample: int = 6) -> str:
-    """Return a stable, bounded signature for request-token correlation logs."""
-
-    if token_ids is None:
-        return "None"
-    if isinstance(token_ids, torch.Tensor):
-        values = token_ids.detach().reshape(-1).cpu().tolist()
-    elif isinstance(token_ids, np.ndarray):
-        values = token_ids.reshape(-1).tolist()
-    else:
-        values = list(token_ids)
-
-    mask = (1 << 63) - 1
-    digest = 1469598103934665603
-    for value in values:
-        item = int(value) & 0xFFFFFFFFFFFFFFFF
-        for shift in (0, 8, 16, 24, 32, 40, 48, 56):
-            digest ^= (item >> shift) & 0xFF
-            digest = (digest * 1099511628211) & mask
-
-    head = values[:sample]
-    tail = values[-sample:] if len(values) > sample else values[:]
-    return f"len={len(values)} hash={digest} head={head} tail={tail}"
-
-
 def _poll_queue_debug_hash(debug_ids: Optional[list[str]]) -> int:
     if not debug_ids:
         return 0