OceanBase Database provides a wide range of features to protect the security and reliability of data in various scenarios where a manual or non-manual exception occurs. These features include multi-replica replication and primary/standby cluster for high availability (HA), as well as recycle bin, flashback query, and backup and restoration.
Hardware exceptions
When a disk becomes faulty, the data stored on the disk may encounter errors. In OceanBase Database, a checksum is stored for REDO logs and the data in SSTables. When data is read, strict checksum verification is performed to detect errors in a timely manner. The mechanism that detects silent data corruption can also detect errors in cold blocks that have not been read for a long period of time. For more information about the mechanism, see Detect silent data corruption.
When a network device such as a switch becomes faulty, a packet error may occur. Strict packet verification and check are performed for all network packets that are exchanged between clients and OBServers and within the OceanBase clusters.
Manual exceptions
Database objects that are deleted by mistake can be restored from the recycle bin.
When an application encounters a software defect, an operation error, or a malicious attack, the table data is regarded as error data. You can use the flashback query feature to read the historical snapshots of the table data and then restore data. You can also use the restore point feature to record data snapshots prior to specific major changes. This way, when an exception occurs, the snapshots can be used to restore data.
In OceanBase Database, the backup and restoration feature provides an efficient mechanism to back up all data and metadata. The feature also serves as the most reliable method for protecting user data. For more information about backup and restoration, see Overview.