CoW systems differ from the other mirroring types. Instead of failing over from primary SDs to secondaries, the administrator copies the contents of an old snapshot to the production SDs. If the P-Vols are lost, S-Vols cannot be used for recovery, because S-Vols are stored as deltas against the P-Vols. Recovery is, therefore, used for recovery from an application-level problem, such as a virus attack or an unwanted database update, rather than to recuperate from physical damage to the storage.
To recover production SDs
- Unmount all file systems on the span.
- Use your storage management interface to copy all the SDs in one snapshot over all the SDs in the span. If the source SDs are not in a single point-in-time snapshot of the span, or if some production SDs are updated and some not, the data will be lost.
- On the server, use the sd-rescan-cod <span-instance-name> command to reload the on-disk configuration, which may have changed.
- Use the mount command to mount the filesystems.