Data integrity after a failure

Data integrity after a failure is dependent on the processing mode (SYNCH, ADAPTIVE-COPY, ASYNCH) and the number of pending writes.

If a link or device error occurs before or during a failure, the source and target units will no longer be synchronized. In other words, if the internal synchronization process is interrupted by a failure in SYNCH or ASYNCH processing mode, the source and target units contain inconsistent data. This is also the case when the synchronization process started by /RESUME-REMOTE-COPY is interrupted.

The situation is even more complicated if not all logically interdependent data is located on the same remote copy pair. In the following case, a device or link error occurs while a DB record is being written. The server does not respond to the error and writes the associated LOG record to another remote copy pair that is not affected by the error. The data on the target units with the DB and LOG records is then inconsistent.

There are two ways of preventing this:

• By using SYNCH processing mode and protecting the target units with RAID1, RAID5 or RAID6. It is advisable to configure the target units with the logically interlinked data on the same storage system. In this case, only a remote link failure can interrupt operation, and this affects all devices at the same time.

• By using SYNCH processing mode and entering /MODIFY-REMOTE-COPY-PARAMETER ...,ON-ERROR=*HOLD for all source units. If a device or link error occurs, server access is interrupted. A decision can then be made as to whether to continue with the remaining unit only (failure of an individual device, not a disaster) or to suspend operation in order to maintain data consistency.