Fault Tolerance

Rivet actors are designed to be fault-tolerant and resilient to various types of failures. This guide explains how actors handle different failure scenarios and how to build reliable applications.

Actor crashes

• When an actor crashes, Rivet automatically starts the actor again (with an exponential backoff).
• The actor's state will still be there when it restarts.
• _onStart will be called when the actor starts again.

Hardware, System, and Network Failures

While uncommon, hardware, system, and network failures can cause an actor to crash. Rivet is designed to handle these scenarios by automatically restarting the actor. (with an exponential backoff).

Actor state

Actor state is automatically restored on startup. Read more here.

Client reconnection on crash

See connection docs.