Adaptive Server disables a mirror only when it encounters an I/O error on a mirrored device. For example, if Adaptive Server tries to write to a bad block on the disk, the resulting error disables mirroring for the device. However, processing continues without interruption on the unaffected mirror.
The following conditions do not disable a mirror:
An unused block on a device is bad. Adaptive Server does not detect an I/O error and disables mirroring until it accesses the bad block.
Data on a device is overwritten. This might happen if a mirrored device is mounted as a UNIX file system, and UNIX overwrites the Adaptive Server data. This causes database corruption, but mirroring is not disabled, since Adaptive Server would not encounter an I/O error.
Incorrect data is written to both the primary and secondary devices.
The file permissions on an active device are changed. Some system administrators may try to test disk mirroring by changing permissions on one device, hoping to trigger I/O failure and unmirror the other device. But the UNIX operating system does not check permissions on a device after opening it, so the I/O failure does not occur until the next time the device is started.
Disk mirroring is not designed to detect or prevent database corruption. Some of the scenarios described can cause corruption, so you should regularly run consistency checks such as dbcc checkalloc and dbcc checkdb on all databases. See Chapter 10, “Checking Database Consistency,” for a discussion of these commands.
