This section describes the fault tolerance mechanisms built into Impala. Fault tolerance allows Impala to continue to operate in the face of hardware or process faults.
Impala uses the statestored process to provide cluster membership for an Impala cluster. This is one of the core ways Impala provides fault tolerance. The statestore periodically sends heartbeats (RPCs) to each impalad process. If an impalad stops responding to heartbeats, the statestored will consider the impalad as failed, and it will remove it from the cluster membership state. The updates to the cluster membership will be broadcasted to all impalads.