Using log extract data sets

Overview

The Recovery utility can write sorted log data to the log extract data set, an external file that resides on DASD.

The following information applies to log extract data sets:

• If a recovery job fails after the log sort task has completed, and if the job uses the Automatic Restart feature and log extract data sets, the job can be restarted with no reprocessing of log data. The incomplete recovery tasks can read the log extract data set at restart. If you use the Automatic Restart feature without log extract data sets, the utility must reread and resort the logs at restart.
• The utility writes an index record at the front of the log extract data set. This index record identifies each data set group (DSG) that is contained in the data set, along with the relative track location (TTR) of the DSG. This index record allows each recovery task to read only the data for the DSG that the task is recovering. When the log extract process has written all of the log extract records for a particular DSG, it notifies the corresponding recovery task that recovery of the DSG can begin.
• Task overlap is improved significantly because the log extract data set can be accessed by many recovery tasks at the same time. Without this data set, the number of truly active recovery tasks is limited by the number of sort tasks that you specify and by the size of internal buffer pools.
• Using the log extract data set can improve performance. Instead of waiting for the recovery task to receive the sorted log records, the sort task can proceed as fast as it can to process the records and write them to the log extract data set. For this processing to occur, you must specify the DIRECT(N) keyword; for more information, see Using-the-DIRECT-keyword.
• After the utility creates the log extract data set, it can terminate the sort task, releasing a significant amount of virtual storage that was occupied by the sort task.
• The utility does not record the data set names of log extract data sets in the RECON. If you are using the Automatic Restart feature, the utility records these data set names in the checkpoint data set. The utility writes log extract data sets in a nonstandard format; they are not compatible with a standard SLDS.
• You can use dynamic allocation models or DD statements in the JCL to allocate log extract data sets. The dynamic allocation model type for a log extract data set is XL.
• The utility can write multiple copies of a log extract data set, but it records only one completed copy in the checkpoint data set. Writing an additional copy can be useful if the first copy experiences a failure (such as an out-of-space condition); if a secondary copy is still valid, the task can continue to write records to it, and recovery tasks that were using it (during the original job step execution, not a restart of the job step) automatically switch from the failed copy to the secondary copy.

Creating log extract data sets

To create log extract data sets, perform the following steps:

1.  Define how to allocate the log extract data set by creating a dynamic allocation model (or by using an existing one) or by defining a DD statement in the job step JCL.You can predefine the dynamic allocation model through the ISPF interface or the Database Administration Console, or you can use the MODEL control statement to define it in the job step. The log extract data set must be allocated on DASD. You can use the &SORT substitution keyword in the model data set name to make the name unique for each sort task.

   For more information, see Using-dynamic-allocation-models.

2. Specify the XLOG control statement in the RVPSYSIN control statement data set. Specify the required keywords as follows:

   • For the NAME keyword value, specify a name to identify this log extract data set.
   • For the DDN keyword value, specify the name of the dynamic allocation model or the ddname of the DD statement from step 1.

   The XLOG control statement must follow the GLBL control statement and any MODEL control statements (that are named on the XLOG statement) and must precede the SORT statement (that names the XLOG statement).

3. Specify the XLOG keyword on the SORT control statement that is used for processing the input log.For the XLOG keyword value, specify the same value as you specified for the NAME keyword on the associated XLOG control statement.

   Example

   The following example demonstrates how to create and use a log extract data set named EXTLOG1 from a dynamic allocation model named EXTMODEL:

   //RVPSYSIN DD *
   ...
   MODEL NAME(EXTMODEL) UNIT(SYSDA) SPACE(C,10,5) ...
   XLOG NAME(EXTLOG1) DDN((*,EXTMODEL))
   SORT SORT(SRT1) XLOG(EXTLOG1)
   ...