change order

Author: richardhuo-nv

src/request_executor.cc

@@ -485,6 +485,9 @@ RequestExecutor::Infer(
           reinterpret_cast<void*>(infer_payload->ResponseAllocUserp().get()),
           InferResponseComplete, reinterpret_cast<void*>(infer_payload.get())));
 
+      THROW_IF_TRITON_ERROR(
+          TRITONSERVER_ServerInferAsync(server_, irequest, trace));
+
       {
         std::lock_guard<std::mutex> lk(on_going_request_addresses_mu_);
         on_going_request_addresses_.insert(
@@ -493,15 +496,9 @@ RequestExecutor::Infer(
       // Store the inference request address submitted to the Triton server for
       // retrieval
       infer_payload->SetRequestAddress(reinterpret_cast<intptr_t>(irequest));
-
-      THROW_IF_TRITON_ERROR(
-          TRITONSERVER_ServerInferAsync(server_, irequest, trace));
     }
   }
   catch (const PythonBackendException& pb_exception) {
-    EraseRequestAddress(reinterpret_cast<intptr_t>(irequest));
-    infer_payload->SetRequestAddress(0L);
-
     LOG_IF_ERROR(
         TRITONSERVER_InferenceRequestDelete(irequest),
         "Failed to delete inference request.");