[HiCache] feat: add multi tenant with prefix tag (#9256)

python/sglang/srt/mem_cache/storage/mooncake_store/mooncake_store.py

@@ -151,6 +151,12 @@ class MooncakeStore(HiCacheStorage):
             )
             per_tp_local_buffer_size = self.config.local_buffer_size // tp_scale_factor
 
+            # Check if extra_backend_tag should be passed to MooncakeDistributedStore
+            self.extra_backend_tag = None
+            if extra_config and "extra_backend_tag" in extra_config:
+                self.extra_backend_tag = extra_config["extra_backend_tag"]
+                logger.info(f"Using extra_backend_tag: {self.extra_backend_tag}")
+
             ret_code = self.store.setup(
                 self.config.local_hostname,
                 self.config.metadata_server,
@@ -257,6 +263,11 @@ class MooncakeStore(HiCacheStorage):
         host_indices: torch.Tensor,
         extra_info: Optional[HiCacheStorageExtraInfo] = None,
     ) -> List[bool]:
+        # Apply extra_backend_tag prefix if available
+        if self.extra_backend_tag is not None:
+            prefix = self.extra_backend_tag
+            keys = [f"{prefix}_{key}" for key in keys]
+
         key_strs, buffer_ptrs, buffer_sizes = self._batch_preprocess(keys, host_indices)
         get_results = self._get_batch_zero_copy_impl(
             key_strs, buffer_ptrs, buffer_sizes
@@ -269,6 +280,11 @@ class MooncakeStore(HiCacheStorage):
         host_indices: torch.Tensor,
         extra_info: Optional[HiCacheStorageExtraInfo] = None,
     ) -> List[bool]:
+        # Apply extra_backend_tag prefix if available
+        if self.extra_backend_tag is not None:
+            prefix = self.extra_backend_tag
+            keys = [f"{prefix}_{key}" for key in keys]
+
         key_strs, buffer_ptrs, buffer_sizes = self._batch_preprocess(keys, host_indices)
         exist_result = self._batch_exist(key_strs)