A continuous backup strategy is required to retain all updated copies of journaled page images over a limited period of time. Point in time recovery requires that all updated page images be retained for the entire window of recovery. Depending on the frequency of updates and the length of the supported recovery window, the amount of recovery journal data can grow to a significant volume. Therefore, the length of the recovery window must reflect a balance between keeping the volume of historical recovery data to a manageable size and providing for a large enough recovery period to be useful. For more information on developing a continuous backup, refer to
Using the Continuous Backup Approach.
This step is necessary to provide a base to continually collect ongoing duplicate page images for point in time recovery. Cleansing the BACKUP file can be accomplished with a special sort using S6BSPX35 or by using the S6BTLSRP (Select Recovery Pages) utility specifying a recovery date of the current runtime.
When a recovery is required, the affected segments must be varied offline or the Data Object Broker terminated with a controlled shutdown. If a DASD failure hits various files, you must recover the Data Object Broker using the procedures for various data set failures. Refer to
Recovering From Non-Page Data Set Failures for more information. After a successful Data Object Broker recovery, vary the segments to be recovered offline and spin the active journals to flush out any pending updates to the segments.
All online segments at the time of termination or error condition are candidates for recovery and should be checked afterwards with the S6BBRPTR (Batch Pointer Check) utility for consistency.
After varying the segments to be recovered offline or successfully shutting down the Data Object Broker, ensure that the journal data sets are spun and consolidated into the backup cycle. This is accomplished by submitting jobs SPIN01, SPIN02, SPINMRG, and BKUPCON.
You must modify the INCLUDE SORT statements in BKUPCON to remove journal records for segments you do not wish to recover. The example below selects pages for Segment 1 only and the resultant file will be a back up of only that segment.
Specify the parameter ‘DATE=YYYYMMMDD,TIME=HHMMSS’ to the S6BTLSRP utility, identifying the time up to which you want to restore your segments. The time selected should match the end of a checkpoint. Use the message S6BKX052L (end checkpoint) produced by the Data Object Broker for determining appropriate timestamps for recovery.
If the EXTRACT file is clean, use it in the S6BTLRPS utility to perform the point in time recovery. Ensure that all pages updated by the Data Object Broker subsequent to the point of recovery are rewritten with previously timestamped updated page images. The recovered segments must not contain any pages updated beyond the recovered point in time. This can best be accomplished by performing a
DELETE and
FORMAT of the affected segments prior to the restore with the S6BTLRPS utility.
The BACKUP file output from step 8 can be used as the basis for the next point in time recovery cycle. You must ensure that no previously journaled page images with timestamps after the time of recovery remain in your continuous backup cycle. If any such page image remains in your backup cycle, page corruptions occur if the BACKUP file is used for subsequent recovery operations.
If the output from step 8 is clean, vary the segments recovered back online to the Data Object Broker. If the Data Object Broker was shut down, restart the Data Object Broker.
