An intermittent failure is a failure that lingers: after one occurrence, the probability of another is elevated. It still resolves automatically, usually after a delay. section 3.3.3
The classification depends on two observable properties:
- Conditional likelihood: the probability of failure F2 after failure F1 is higher than the probability of F2 occurring independently. section 3.3.3
- Resolution: the failure autorepairs without intervention, but recovery may require time for the underlying condition to change. section 3.3.3
- A failed connection caused by an outdated routing table is intermittent: connection attempts may continue failing until the router updates its table. section 3.3.3
- An extended employee absence provides a distributed-systems analogy: the component is unavailable for a while, then returns without permanent loss. section 1.6
Figure 3.13 Intermittent failure
Use the temporal classification diagram to contrast intermittent failures with transient and permanent failures. section 3.3.3
