Disclosed herein are system, method, and computer program product embodiments for fault tolerant communications in a distributed system. An embodiment operates by determining a failure of a first node of a cluster of a plurality of nodes. Prior to the failure, a second node transmitted a first message to the first node. The second node determines content corresponding to the first message. A third node to which to provide the content of the first sub-task is identified. A second message, including content corresponding to the first message, is transmit to the third node.