Preserve CP shared-KV cache-hit diagnostics

Warm-cache GSM8K failures needed request-to-log correlation across scheduler, prefill transfer, and Mooncake CP filtering. The added diagnostics are gated by existing CP shared-KV debug envs and include bounded token signatures plus transfer page summaries so future debugging can identify whether a failed request hit L1, L2, or transfer truncation paths.\n\nThe findings document records the ruled-out hypotheses and the RAGGED current-row contract failure, preventing repeated log archaeology after context compaction.\n\nConstraint: Production hot paths must not emit these logs unless SGLANG_CP_SHARED_KV_BS_GT1_DEBUG or existing CP shared-KV debug gates are enabled.\nRejected: Add a new debug env | reuse of existing bs>1 debug gate avoids more runtime switches.\nRejected: Store full token ids in logs | bounded signatures are enough for correlation without huge logs.\nConfidence: medium\nScope-risk: narrow\nDirective: Keep these diagnostics gated and bounded; do not convert them to unconditional INFO logs.\nTested: Remote container py_compile as part of the RAGGED fix validation.\nNot-tested: Long-running production log-volume impact with debug enabled.

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

@@ -64,6 +64,21 @@ 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()
 
@@ -1211,6 +1226,19 @@ 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",
@@ -1326,6 +1354,18 @@ 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,
@@ -1706,6 +1746,22 @@ 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(