Using database change accumulation input for recovery

You must include any log volumes recorded since the last time you ran the Database Change Accumulation utility as input to the recovery—either in their current format or after being passed through the Database Change Accumulation utility.

When you run the Database Change Accumulation utility, you can specify the DBDSs for which you want to accumulate change records. For each DBDS, you can either ignore change records or write them to the optional output log data set. By specifying a subset of the DBDSs (called a change accumulation group), you can reduce the size of the change accumulation data set, and improve performance of the Database Recovery utility. However, you also must run the Database Change Accumulation utility (and pass the log data set) for each change accumulation group.

If you select only one DBDS each time you run the Database Change Accumulation utility, you must run the utility (and pass the log data set) once for each DBDS. With this technique, each output change accumulation file contains only those records needed for the recovery of a single DBDS.

You might improve the performance of the Database Change Accumulation utility by performing frequent quiesces of your databases. Quiescing a database creates a point of consistency across databases and logs which reduces the number of log records that need to be accumulated. When creating a point of consistency, new updates are held and all in-progress updates are committed and written to DASD. After the last in-progress update is written to DASD a deallocation (DEALLOC) time stamp is added to the database allocation (ALLOC) record in the RECON data set to create a new recovery point and updates to the database resume.

The following are some common recovery techniques:

• Define a single change accumulation group for all DBDSs.

  In this case, the Database Change Accumulation utility reads the log data set only once, but the Database Recovery utility must read records in the output change accumulation data set for potentially unnecessary databases. This technique improves the performance of the Database Change Accumulation utility, but degrades performance of the database recovery.

• Define several change accumulation groups.

  In this case, you divide your databases into groups of several databases, often by volume or by application. If you take image copies frequently, you can keep the number of records scanned by the Database Recovery utility to a minimum.

• Define no change accumulation groups. Run the Database Change Accumulation utility, but not as a regular job.

  When you need to recover a database, specify that the utility accumulate changes for the affected DBDS only. Provide all log data sets created since the latest image copy of the data set to be recovered. If you use one of the image copy utilities, you must include the log created when the image copy was made as input to the Database Change Accumulation utility.

  If you are considering this approach, you should make some test runs to see whether the sum of the execution times of the Database Change Accumulation and Database Recovery utilities is less than the time needed to run the Database Recovery utility with unaccumulated log data sets.