diff --git a/pytorch_lightning/trainer/data_loading.py b/pytorch_lightning/trainer/data_loading.py
index ee545128..43e87928 100644
--- a/pytorch_lightning/trainer/data_loading.py
+++ b/pytorch_lightning/trainer/data_loading.py
@@ -233,7 +233,7 @@ class TrainerDataLoadingMixin(ABC):
                 self.get_val_dataloaders()
 
             # wait for all processes to catch up
-            torch_xla.core.xla_model.rendezvous()
+            torch_xla.core.xla_model.rendezvous("pl.TrainerDataLoadingMixin.get_dataloaders")
 
         # support IterableDataset for train data
         self.is_iterable_train_dataloader = (
diff --git a/pytorch_lightning/trainer/trainer.py b/pytorch_lightning/trainer/trainer.py
index 37c59a11..4d204bd2 100644
--- a/pytorch_lightning/trainer/trainer.py
+++ b/pytorch_lightning/trainer/trainer.py
@@ -1002,7 +1002,7 @@ class Trainer(TrainerIOMixin,
         # wait for all models to restore weights
         if self.on_tpu and XLA_AVAILABLE:
             # wait for all processes to catch up
-            torch_xla.core.xla_model.rendezvous()
+            torch_xla.core.xla_model.rendezvous("pl.Trainer.run_pretrain_routine")
 
         # set up checkpoint callback
         self.configure_checkpoint_callback()
diff --git a/pytorch_lightning/trainer/training_io.py b/pytorch_lightning/trainer/training_io.py
index 1bc14966..569c5718 100644
--- a/pytorch_lightning/trainer/training_io.py
+++ b/pytorch_lightning/trainer/training_io.py
@@ -183,7 +183,7 @@ class TrainerIOMixin(ABC):
         # wait for all models to restore weights
         if self.on_tpu and XLA_AVAILABLE:
             # wait for all processes to catch up
-            torch_xla.core.xla_model.rendezvous()
+            torch_xla.core.xla_model.rendezvous("pl.TrainerIOMixin.restore_weights")
 
         # clear cache after restore
         if self.on_gpu: