Nonrecoverable conditions

• Failure of both mirrored drives.

• Failure of both split-volume set halves.

• Failure of the source drive during the repair operation.

• Drive failure during join.

If both mirrored drives fail

When a single drive of a pair fails, a period of time exists while the drive is replaced and the new one is repaired. If the remaining drive fails during this time, the failure is identical to a disk drive failure on a non-mirrored system.

If both split volume set halves fail

If both the user volumes and backup volumes suffer a disk failure, the situation is nonrecoverable. This applies even if the volumes were not previously mirrored partners (same copies), but different members of the volume set. Either the user volumes or backup volumes must be complete in order for the volume set to return to a known consistent point. User volumes and backup volumes may not be mixed to create a volume set. The split-volume set cannot be joined, and must be recreated.

If a drive fails during the repair operation

Whenever there is an unexpected failure that necessitates a rebooting of the system, or when a split-volume set is joined into a mirrored volume set, all mirrored pairs undergo a repair process. This is necessary to guarantee the consistency of the disks. During the repair operation, one disk is copied to the other. If the source disk fails during this period, there is no way to recover from it without restoring the files from tape.

If the source drive of a staged or repairing volume pair becomes unresponsive (due to loss of power, for example) or suffers a hard error, the system behaves just as if there were no mirroring present. If the destination drive of a repairing volume pair becomes unresponsive or suffers a hard error, it is marked DISABLED. If the destination drive of a staged volume pair becomes unresponsive after being staged, but before transitioning to be repaired, it is marked DISABLED at repair time.

If a drive fails during the join operation

This is a more specific case of a failure during repair. If a JOINMIRRSET completes and during the repair one of the REPAIR-SRCE disks becomes disbaled, the volume set cannot be recovered and has to be reloaded.

Additionally, the system cannot recover from the following user errors:

If you misuse the SUSPENDMIRRVOL command

It is possible through the use of the SUSPENDMIRRVOL command to force the system to mount and use a drive that is bad. Because of drive errors, it may not have been possible to mark the drive itself as bad when the errors occurred. Now if the volume set is remounted without the good partner, there may be no way for the system to know that the drive is bad. In this case the use of the SUSPENDMIRRVOL command could force the mounting of the bad drive, which may contain data that has not been updated. While this situation is highly unlikely, care must be taken when using this command to ensure that the disk in question is good.

If you create a software error

A mirrored disk system offers no protection against software errors that mistakenly write bad data to disk. The bad data is mirrored just as any other writing to disk.

If you modify a disk

Privileged mode users could, through DEBUG, make modification to one partner of a mirrored pair without notifying the mirrored disk system. This causes the partner disks to be inconsistent and could lead to unpredictable results.