Hicache L3 backend mooncake optimization configuration reading method (#10319)

Co-authored-by: Teng Ma <sima.mt@alibaba-inc.com>
Co-authored-by: huangtingwei <141888744+huangtingwei9988@users.noreply.github.com>
Co-authored-by: shicang <shicang@shicang>
Co-authored-by: Shangming Cai <csmthu@gmail.com>

python/sglang/srt/mem_cache/storage/mooncake_store/README.md

@@ -119,7 +119,10 @@ Note: If `MOONCAKE_GLOBAL_SEGMENT_SIZE` is set to a non-zero value when starting
 
 Mooncake configuration can be provided via environment variables. Note that, for optimal performance, the Mooncake backend currently supports only the `page_first` layout (which optimizes memory access patterns for KV cache operations).
 
-There are two ways to configure Mooncake: 1. Using environment variables; 2. Using extra-config of sglang arguments.
+There are three ways to prepare mooncakes:
+1. Use environment variables;
+2. Use json configuration files;
+3. Additional configuration using the sglang parameter.
 
 **Using env variables to configure Mooncake**
 
@@ -143,6 +146,21 @@ Parameter Explanation:
 * `MOONCAKE_DEVICE`: The RDMA devices used by Mooncake. This parameter is required only when the protocol is set to `"rdma"`. Available devices can be listed using the `ibv_devices` command.
 * `MOONCAKE_GLOBAL_SEGMENT_SIZE`: The amount of memory (in bytes) contributed to the global memory pool. If at least one `store service` is launched, then this value could be set to `0`. In this case, the `SGLang server` will not contribute any memory to the system. Note that KV tensors cached in the contributed memory will be lost once this process terminates; however, this will not cause any system errors.
 
+**Using JSON file to configure Mooncake**
+
+```bash
+export SGLANG_HICACHE_MOONCAKE_CONFIG_PATH=/sgl-workspace/sglang/benchmark/hicache/mooncake_config.json
+echo '{
+    "local_hostname": "localhost",
+    "metadata_server": "http://localhost:8080/metadata",
+    "master_server_address": "localhost:50051",
+    "protocol": "rdma",
+    "device_name": "mlx5_0,mlx5_1",
+    "global_segment_size": 2684354560,
+    "local_buffer_size": 0
+}' > ${SGLANG_HICACHE_MOONCAKE_CONFIG_PATH}
+```
+
 **Using extra-config of sglang arguments to configure Mooncake**
 
 ```bash

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

@@ -11,7 +11,7 @@ from sglang.srt.mem_cache.hicache_storage import HiCacheStorage, HiCacheStorageC
 
 DEFAULT_GLOBAL_SEGMENT_SIZE = 4 * 1024 * 1024 * 1024  # 4 GiB
 DEFAULT_LOCAL_BUFFER_SIZE = 16 * 1024 * 1024  # 16 MB
-
+DEFAULT_MOONCAKE_CONFIG_PATH_ENV = "SGLANG_HICACHE_MOONCAKE_CONFIG_PATH"
 logger = logging.getLogger(__name__)
 
 
@@ -28,13 +28,13 @@ class MooncakeStoreConfig:
     @staticmethod
     def from_file() -> "MooncakeStoreConfig":
         """Load the config from a JSON file."""
-        file_path = os.getenv("MOONCAKE_CONFIG_PATH")
-        if file_path is None:
-            raise ValueError(
-                "The environment variable 'MOONCAKE_CONFIG_PATH' is not set."
-            )
-        with open(file_path) as fin:
-            config = json.load(fin)
+        file_path = os.getenv(DEFAULT_MOONCAKE_CONFIG_PATH_ENV)
+        try:
+            with open(file_path) as fin:
+                config = json.load(fin)
+        except Exception as e:
+            raise RuntimeError(f"Failed to load config from {file_path}: {str(e)}")
+
         return MooncakeStoreConfig(
             local_hostname=config.get("local_hostname"),
             metadata_server=config.get("metadata_server"),
@@ -101,6 +101,7 @@ class MooncakeStoreConfig:
 
 
 class MooncakeStore(HiCacheStorage):
+
     def __init__(self, storage_config: HiCacheStorageConfig = None):
         try:
             from mooncake.store import MooncakeDistributedStore
@@ -129,6 +130,10 @@ class MooncakeStore(HiCacheStorage):
                 logger.info(
                     "Mooncake Configuration loaded from extra_config successfully."
                 )
+            elif os.getenv(DEFAULT_MOONCAKE_CONFIG_PATH_ENV):
+                # Load from config file
+                self.config = MooncakeStoreConfig.from_file()
+                logger.info("Mooncake Configuration loaded from file successfully.")
             else:
                 # Load from environment variables
                 self.config = MooncakeStoreConfig.load_from_env()