E-MapReduce (EMR) provides the best-effort delivery and automatic compensation policies to improve the reliability, availability, and stability of an EMR cluster. You can choose both the policies or either one of them to implement fault tolerance for nodes in the EMR cluster. If you turn on Best-effort Delivery for a cluster during a scale-out activity, the system adds all successfully created nodes to the cluster in a best-effort manner to ensure continuity of the services that are deployed in the cluster even specific nodes failed to be created. If you turn on Automatic Compensation for an existing node group or turn on Automatic Compensation when you add a specific node group, after an exception occurs on a node, the system automatically adds a node and replaces the abnormal node with the added node.