Читать книгу From Traditional Fault Tolerance to Blockchain - Wenbing Zhao - Страница 65

The sender-based message logging protocol described above ensures proper recovery of a distributed system as long as a single failure occurs at a time. That is, after a process fails, no other processes fail until the failed process is fully recovered. Note that the protocol cannot cope with two or more concurrent failures. If two or more failures occur concurrently, the determinant for some regular messages (i.e., the rsn values) might be lost, which would lead to orphan processes and the cascading rollback (i.e., the domino effect).