7.5.2.2 Host State Detection

In practice it turned out to be crucial to detect hosts which are temporarily not reachable. Therefore, a host object considers its corresponding host unreachable or unavailable, if it did not report its load between two subsequent calls of the :trigger-spy method. As soon as the host is resuming to report its load it is considered to be available again. This way of host state detection results in a very stable job farming system since even disastrous situations can be managed and attempts to use corrupted hosts can successfully be prevented. Moreover, as soon as a host, or the network it is connected to, recovers, it is employed for computation.