The node which runs the application code on the Spark cluster is Spark worker node. These are the slave nodes. Any of the worker nodes running executor can fail, thus resulting in loss of in-memory If any receivers were running on failed nodes, then their buffer data will be lost.