Minor fix (requested by HQ)

mindspore/ccsrc/minddata/dataset/engine/cache/cache_service.cc

@@ -39,8 +39,10 @@ Status CacheService::DoServiceStart() {
   if (cache_mem_sz_ > 0) {
     auto avail_mem = CacheServerHW::GetTotalSystemMemory();
     if (cache_mem_sz_ > avail_mem) {
-      // Output a warning that we use more than recommended. If we fail to allocate, we will fail anyway.
-      MS_LOG(WARNING) << "Requesting cache size " << cache_mem_sz_ << " while available system memory " << avail_mem;
+      // Return an error if we use more than recommended memory.
+      std::string errMsg = "Requesting cache size " + std::to_string(cache_mem_sz_) +
+                           " while available system memory " + std::to_string(avail_mem);
+      return Status(StatusCode::kOutOfMemory, __LINE__, __FILE__, errMsg);
     }
     memory_cap_ratio = static_cast<float>(cache_mem_sz_) / avail_mem;
   }

mindspore/ccsrc/minddata/dataset/engine/datasetops/cache_base_op.cc

@@ -308,6 +308,7 @@ Status CacheBase::Prefetcher(int32_t worker_id) {
           // If we get some network error, we will attempt some retries
           retry_count++;
         } else if (rc.IsError()) {
+          MS_LOG(WARNING) << rc.ToString();
           return rc;
         }
       } while (rc.IsNetWorkError());

mindspore/ccsrc/minddata/dataset/engine/ir/cache/dataset_cache_impl.h

@@ -32,7 +32,8 @@ class DatasetCacheImpl : public DatasetCache {
   ///
   /// \brief Constructor
   /// \param id A user assigned session id for the current pipeline.
-  /// \param mem_sz Size of the memory set aside for the row caching (default=0 which means unlimited).
+  /// \param mem_sz Size of the memory set aside for the row caching (default=0 which means unlimited,
+  ///     note that it might bring in the risk of running out of memory on the machine).
   /// \param spill Spill to disk if out of memory (default=False).
   /// \param hostname optional host name (default="127.0.0.1").
   /// \param port optional port (default=50052).

mindspore/ccsrc/minddata/dataset/include/datasets.h

@@ -1011,7 +1011,8 @@ std::shared_ptr<VOCDataset> VOC(const std::string &dataset_dir, const std::strin
 
 /// \brief Function the create a cache to be attached to a dataset
 /// \param id A user assigned session id for the current pipeline.
-/// \param mem_sz Size of the memory set aside for the row caching (default=0 which means unlimited).
+/// \param mem_sz Size of the memory set aside for the row caching (default=0 which means unlimited,
+///     note that it might bring in the risk of running out of memory on the machine).
 /// \param spill Spill to disk if out of memory (default=False).
 /// \param hostname optional host name (default="127.0.0.1").
 /// \param port optional port (default=50052).

mindspore/dataset/engine/cache_client.py

@@ -30,7 +30,8 @@ class DatasetCache:
 
     Args:
         session_id (int): A user assigned session id for the current pipeline.
-        size (int, optional): Size of the memory set aside for the row caching (default=0 which means unlimited).
+        size (int, optional): Size of the memory set aside for the row caching (default=0 which means unlimited,
+            note that it might bring in the risk of running out of memory on the machine).
         spilling (bool, optional): Whether or not spilling to disk if out of memory (default=False).
         hostname (str, optional): Host name (default="127.0.0.1").
         port (int, optional): Port to connect to server (default=50052).