While fault containment might partially or even completely mask failures from the system's user, the system itself must always be aware of the failures.
Both transient and non-transient faults must be detected and reported to the system's monitoring tools. A high number of transient faults on a certain system component may indicate overload and the need to increase capacity. Non-transient, repeatable failures indicate a code or configuration defect that needs to be corrected. Even when a failure is caused by a service external to the system, it is still better to adjust the system's behavior to prevent the error, instead of repeatedly swallowing exceptions. Fault detection is necessary to identify and correct the system defects ...
