[Feature] ClusterExecutor: Store file errors in output files (#918)

Author: jan-janssen

src/executorlib/backend/cache_parallel.py

@@ -39,11 +39,11 @@ def main() -> None:
 
     time_start = time.time()
     apply_dict = {}
-    if mpi_rank_zero:
-        apply_dict = backend_load_file(file_name=file_name)
-    apply_dict = MPI.COMM_WORLD.bcast(apply_dict, root=0)
-    output = apply_dict["fn"].__call__(*apply_dict["args"], **apply_dict["kwargs"])
     try:
+        if mpi_rank_zero:
+            apply_dict = backend_load_file(file_name=file_name)
+        apply_dict = MPI.COMM_WORLD.bcast(apply_dict, root=0)
+        output = apply_dict["fn"].__call__(*apply_dict["args"], **apply_dict["kwargs"])
         result = (
             MPI.COMM_WORLD.gather(output, root=0) if mpi_size_larger_one else output
         )

src/executorlib/task_scheduler/file/backend.py

@@ -68,9 +68,9 @@ def backend_execute_task_in_file(file_name: str) -> None:
     Returns:
         None
     """
-    apply_dict = backend_load_file(file_name=file_name)
-    time_start = time.time()
     try:
+        apply_dict = backend_load_file(file_name=file_name)
+        time_start = time.time()
         result = {
             "result": apply_dict["fn"].__call__(
                 *apply_dict["args"], **apply_dict["kwargs"]