IBM DS8000 User Manual page 185

For this scenario, assume that Site A is operational and that connectivity from Site
B to Site A is available. Use this procedure to restart your production environment
using the volumes from Site B. After the failover operation processes, the volumes
at Site B are available to be mounted and used by a host. You can use the volumes
from Site B to restart your application.
The failback operation resynchronizes the volumes in the following manner
depending on the volume state:
v If a volume at Site A is in a simplex state, all of the data for that volume is sent
from Site B to Site A.
v If a volume at Site A is in a full-duplex or suspended state and without changed
tracks, only the modified data on the volume at Site B is sent to the volume at
Site A.
v If a volume at Site A is in a suspended state but has tracks that have been
modified, the volume at Site B discovers which tracks were modified at any site
and sends both the tracks that were changed on Site A and the tracks that were
marked at Site B from Site A to Site B.
The following assumptions are made for this scenario:
v Paths from Site B to Site A are created.
v Remote mirror and copy volume pairs are created. The Site B volume is the
source volume of the failback operation. This volume was initially the target
volume of the relationship.
The following steps summarize the actions that you must take to return production
to your local site after initiating a planned outage.
1. Perform a failback operation at Site B. As a result, data that has been modified
at Site B for the specified volumes is sent from Site B to Site A. Data that was
modified at Site A after the switch to Site B is overwritten by the data from Site
B.
2. Quiesce your application (still updating volumes at Site B) to cease all write
I/O from updating the source volumes before the application returns to normal
operations. Depending on the host operating system, it might be necessary to
unmount the source volumes.
3. Perform another failover recovery operation at Site A. This process converts the
full-duplex target volumes at Site A to suspended source volumes. The volumes
at Site A start the change recording process while in failover mode.
4. Rescan (depending on your operating system) the fibre-channel devices and
mount the new source volumes at Site A.
5. Restart the applications in normal production at Site A and check for
consistency after the failover operation completes. When your applications are
started, all write I/O to the new source volumes are tracked by the bitmap that
is created.
6. From Site A, create paths in the opposite direction from Site A to Site B.
7. From Site A, perform another failover and failback operation using the source
volumes. This process resynchronizes the volumes between your Sites A and B.
Start your applications at Site A and check for consistency.
8. Mount your volumes at Site A and start your applications on your primary
server.
Chapter 10. Disaster recovery using Copy Services
157