Skip to main content

Failure

The Failure system message is emitted by the runtime when a child actor crashes and its supervisor needs to decide how to recover. The message contains:

  • the failing child's PID,
  • the RestartStatistics gathered for the actor, and
  • the exception or error that triggered the failure.

Supervisors do not receive Failure through their regular mailbox. Instead, the runtime delivers it through the supervision channel so the parent can apply its supervision strategy.

Typical supervisory logic reacts to Failure by selecting one of the built-in directives:

  • ResumeDirective to continue processing without restarting,
  • RestartDirective or RestartChildrenDirective to restart the child,
  • StopDirective to stop the child permanently, or
  • EscalateDirective to propagate the error further up the hierarchy.

Example: logging crashes in .NET​

public override SupervisorDirective HandleFailure(
    ISupervisor supervisor,
    PID child,
    RestartStatistics rs,
    Exception reason)
{
    _logger.LogError(reason, "{Actor} failed after {RestartCount} restarts", child, rs.FailureCount);
    return SupervisorDirective.Restart;
}

See the full supervisor samples in the Proto.Actor .NET examples.

Example: inspecting failures in Go​

func (s *loggingStrategy) HandleFailure(supervisor actor.Supervisor, child *actor.PID, rs *actor.RestartStatistics, reason interface{}) {
    log.Printf("actor %s failed %d time(s): %v", child, rs.FailureCount(), reason)
    supervisor.RestartChildren(child)
}

You can explore more Go supervisors in the Proto.Actor Go examples.