diff --git a/QueryEngine/ResultSetConversion.cpp b/QueryEngine/ResultSetConversion.cpp
index 9c24f4a235..8c667a25f2 100644
--- a/QueryEngine/ResultSetConversion.cpp
+++ b/QueryEngine/ResultSetConversion.cpp
@@ -300,6 +300,7 @@ std::vector<std::shared_ptr<Array>> generate_columns(
   CHECK_GT(col_count, 0);
   std::vector<std::shared_ptr<arrow::Array>> columns(col_count, nullptr);
   auto generate = [&](const size_t i, std::shared_ptr<arrow::Array>& column) {
+    CHECK(column_values[i]);
     column = generate_column(*fields[i], null_bitmaps[i], *column_values[i]);
   };
   if (col_count > 1) {
@@ -441,6 +442,9 @@ void print_serialized_schema(const uint8_t* data, const size_t length) {
 
 std::shared_ptr<PoolBuffer> serialize_arrow_records(const RecordBatch& rb, ipc::FileBlock& block) {
   int64_t rb_sz = 0;
+  if (!rb.num_rows()) {
+    return std::make_shared<PoolBuffer>();
+  }
   ASSERT_OK(ipc::GetRecordBatchSize(rb, &rb_sz));
   auto pool = default_memory_pool();
   auto buffer = std::make_shared<PoolBuffer>(pool);
@@ -464,6 +468,13 @@ void print_serialized_records(const uint8_t* data,
                               const size_t length,
                               const ipc::FileBlock& block,
                               const std::shared_ptr<Schema>& schema) {
+  if (data == nullptr || !length) {
+    std::cout << "No row found" << std::endl;
+    return;
+  }
+
+  CHECK_GT(block.metadata_length, int32_t(0));
+  CHECK_GT(block.body_length, int64_t(0));
   const auto payload = std::make_shared<arrow::Buffer>(data, length);
   auto buffer_reader = std::make_shared<io::BufferReader>(payload);
 
@@ -650,7 +661,9 @@ std::pair<std::vector<std::shared_ptr<arrow::Array>>, size_t> ResultSet::getArro
     row_count = fetch(column_values, null_bitmaps, size_t(0), entry_count);
   }
 
-  return {generate_columns(fields, null_bitmaps, column_values), row_count};
+  return {row_count > 0 ? generate_columns(fields, null_bitmaps, column_values)
+                        : std::vector<std::shared_ptr<arrow::Array>>(),
+          row_count};
 }
 
 arrow::RecordBatch ResultSet::convertToArrow(const std::vector<std::string>& col_names,
@@ -702,7 +715,6 @@ ArrowResult ResultSet::getArrowCopyOnCpu(Data_Namespace::DataMgr* data_mgr,
   arrow::ipc::FileBlock block{0, 0, 0};
   auto serialized_records = arrow::serialize_arrow_records(arrow_copy, block);
   const auto record_key = arrow::get_and_copy_to_shm(serialized_records);
-  CHECK(record_key != IPC_PRIVATE);
   std::vector<char> record_handle_buffer(sizeof(key_t), 0);
   memcpy(&record_handle_buffer[0], reinterpret_cast<const unsigned char*>(&record_key), sizeof(key_t));
 
@@ -733,22 +745,23 @@ ArrowResult ResultSet::getArrowCopyOnGpu(Data_Namespace::DataMgr* data_mgr,
 #ifdef HAVE_CUDA
   arrow::ipc::FileBlock block{0, 0, 0};
   auto serialized_records = arrow::serialize_arrow_records(arrow_copy, block);
-  CHECK(data_mgr && data_mgr->cudaMgr_);
-  auto dev_ptr =
-      reinterpret_cast<CUdeviceptr>(data_mgr->cudaMgr_->allocateDeviceMem(serialized_records->size(), device_id));
-  CUipcMemHandle record_handle;
-  cuIpcGetMemHandle(&record_handle, dev_ptr);
-  data_mgr->cudaMgr_->copyHostToDevice(reinterpret_cast<int8_t*>(dev_ptr),
-                                       reinterpret_cast<const int8_t*>(serialized_records->data()),
-                                       serialized_records->size(),
-                                       device_id);
-  std::vector<char> record_handle_buffer(sizeof(record_handle), 0);
-  memcpy(&record_handle_buffer[0], reinterpret_cast<unsigned char*>(&record_handle), sizeof(CUipcMemHandle));
-
-  return {schema_handle_buffer, serialized_schema->size(), record_handle_buffer, serialized_records->size()};
-#else
-  return {schema_handle_buffer, serialized_schema->size(), {}, 0};
+  if (serialized_records->size()) {
+    CHECK(data_mgr && data_mgr->cudaMgr_);
+    auto dev_ptr =
+        reinterpret_cast<CUdeviceptr>(data_mgr->cudaMgr_->allocateDeviceMem(serialized_records->size(), device_id));
+    CUipcMemHandle record_handle;
+    cuIpcGetMemHandle(&record_handle, dev_ptr);
+    data_mgr->cudaMgr_->copyHostToDevice(reinterpret_cast<int8_t*>(dev_ptr),
+                                         reinterpret_cast<const int8_t*>(serialized_records->data()),
+                                         serialized_records->size(),
+                                         device_id);
+    std::vector<char> record_handle_buffer(sizeof(record_handle), 0);
+    memcpy(&record_handle_buffer[0], reinterpret_cast<unsigned char*>(&record_handle), sizeof(CUipcMemHandle));
+
+    return {schema_handle_buffer, serialized_schema->size(), record_handle_buffer, serialized_records->size()};
+  }
 #endif
+  return {schema_handle_buffer, serialized_schema->size(), {}, 0};
 }
 
 ArrowResult ResultSet::getArrowCopy(Data_Namespace::DataMgr* data_mgr,