mover barrier into write_mesh -- to avoid fs races out of the box

LLNL · Sep 16, 2020 · 03a8956 · 03a8956
1 parent 043da54
commit 03a8956
Show file tree

Hide file tree

Showing 2 changed files with 9 additions and 8 deletions.
diff --git a/src/libs/relay/conduit_relay_io_blueprint.cpp b/src/libs/relay/conduit_relay_io_blueprint.cpp
@@ -1223,6 +1223,15 @@ void write_mesh(const Node &mesh,
         hnd.write(root);
         hnd.close();
     }
+
+    // barrier at end of work to avoid file system race
+    // (non root task could write the root file in write_mesh, 
+    // but root task is always the one to read the root file
+    // in read_mesh.
+
+    #ifdef CONDUIT_RELAY_IO_MPI_ENABLED
+        MPI_Barrier(mpi_comm);
+    #endif
 }
 
 //-----------------------------------------------------------------------------

diff --git a/src/tests/blueprint/t_blueprint_mpi_mesh_relay.cpp b/src/tests/blueprint/t_blueprint_mpi_mesh_relay.cpp
@@ -117,7 +117,6 @@ TEST(blueprint_mpi_relay, basic_use)
                                                   output_base,
                                                   "hdf5",
                                                   comm);
-    MPI_Barrier(MPI_COMM_WORLD);
 
     // read this back using read_mesh, should diff clean
     string output_root = output_base + ".cycle_000000.root";
@@ -168,7 +167,6 @@ TEST(blueprint_mpi_relay, mpi_mesh_examples_braid)
                                                   output_base,
                                                   "hdf5",
                                                   comm);
-    MPI_Barrier(comm);
 
     // read this back using read_mesh, should diff clean
     string output_root = output_base + ".cycle_000000.root";
@@ -243,7 +241,6 @@ TEST(blueprint_mpi_relay, mpi_mesh_examples_spiral_5doms)
                                                   output_base,
                                                   "hdf5",
                                                   comm);
-    MPI_Barrier(comm);
 
     // read this back using read_mesh, should diff clean
     string output_root = output_base + ".cycle_000000.root";
@@ -336,7 +333,6 @@ TEST(blueprint_mpi_relay, mpi_mesh_examples_spiral_1dom)
                                                   "hdf5",
                                                   opts,
                                                   comm);
-    MPI_Barrier(comm);
 
     // read this back using read_mesh, should diff clean
     string output_root = output_base + ".cycle_000000.root";
@@ -467,8 +463,6 @@ TEST(blueprint_mpi_relay, spiral_multi_file)
                                                       opts,
                                                       comm);
 
-        MPI_Barrier(comm);
-
         // count the files
         //  file_%06llu.{protocol}:/domain_%06llu/...
         int nfiles_to_check = nfiles;
@@ -500,8 +494,6 @@ TEST(blueprint_mpi_relay, spiral_multi_file)
             std::cout << " checking: " << fcheck << std::endl;
             EXPECT_TRUE(conduit::utils::is_file(fcheck));
         }
-
-        MPI_Barrier(comm);
     }
 
     // read this back using read_mesh