Restarting a BMC AMI Recover job

The value that you specify for the restart parameter (Restart parameter (restartParm)) controls the restarting of the utility. You can restart the utility from the last recorded checkpoint by specifying NEW/RESTART or RESTART for the restart parameter value.

If a table space, index, or index space was marked unrecoverable because of unavailable recovery resources, BMC AMI Recover does not attempt to recover that object in a restarted job. If the recovery resources have been made available since the original job, BMC AMI Recover recovers the object by using the NEW restart parameter.

The following table shows the kinds of restart that are possible for each phase of BMC AMI Recover:

To restart a failed job successfully, you must use the same utility ID as in the original run. Also, you must ensure that the commands in the SYSIN data set are compatible with those in the original job. For example, you can remove objects (table spaces and indexes) from the job, but you cannot add objects, nor can you change critical attributes of the job or its objects. The values for ANALYZE must remain the same. None of the index commands can be changed to another type. For example, REBUILD INDEX cannot be changed to RECOVER UNLOADKEYS. Ensure that none of the objects involved in the recovery have been altered between the time the original job terminated and the time of the restart.

The following must apply in the original and restarted jobs:

• For each RECOVER TABLESPACE, RECOVER INDEXSPACE, and RECOVER INDEX command in the job, LOGAPPLY ONLY, LOGONLY, FROMRBA/FROMLOGPOINT, TORBA/TOLOGPOINT, TOCOPY, and INDEP OUTSPACE (and the associated data sets) must remain the same.

• For each REBUILD INDEX command, the values of WORKDDN (or NOWORKDDN) and INDEP OUTSPACE (and the associated data sets) must remain the same.

• For RECOVER UNLOADKEYS commands, the values of PART and SKEYDDN must remain the same.

• For the RECOVER BUILDINDEX commands, the value of SKEYDDN must remain the same.

To ensure restartability during the BUILD phase for rebuilding an index, specify DISP=(MOD,CATLG,CATLG) in the DD statement for each key work data set. If you must restart the run, the key work data set will be referred to as DISP=OLD by the program. This DISP value results in correct handling during restart. A subsequent step can delete the work data sets conditionally.

Restart considerations when creating output copies

The way in which your restart jobs are set up depends on whether you restart your BMC AMI Recover jobs from the beginning or restart from the point of failure. 

Restart from the beginning

If you plan to always restart failed BMC AMI Recover jobs from the beginning, use the NEW restart parameter in the original job. If you allocate the output copy data sets in the JCL, you must also perform the following things:

• Use DISP=(NEW, CATLG, CATLG) or DISP=(NEW, KEEP, KEEP).

• Use unique data set names for each execution.

  GDGs are helpful for accomplishing this task.

• Code BLKSIZE=0 for disk data sets so that unopened data sets can be migrated successfully, if necessary.

This method leaves empty, unused data sets if disk copies are made for any object not recovered by the failing run.

The advantage of this method is that only the data set names must be modified for the restart. If GDGs are used, however, no modification of the JCL is necessary. Do not use this method under a restart package that modifies the data set names or dispositions on restart. 

Restart from the point of failure

If you plan to always restart failed BMC AMI Recover jobs from the point of failure, use the NEW/RESTART restart parameter in the original job. If you allocate copy data sets in the JCL, you must also perform the following actions:

• Use DISP=(MOD, CATLG, CATLG) or DISP=(NEW, KEEP, KEEP).

• Perform the following additional steps at restart:

  • For stacked tape copies that use GDGs, modify the data set names to indicate the generation relative to the restart. Modify the (+ n) value to (+ n- m), where m is the relative generation number for the last cataloged generation in the original execution.

  • For cataloged stacked tape copies, remove VOL=REF= from the copy data set DD statements for the copy creation that failed. This removal tells the system to use the catalog for volume information.

  • Failure to remove VOL=REF= causes the restarted data set to get a 'not cataloged' message and causes a multiple volume data set to be on a different set of volumes than the original, cataloged data set. If the restarted copy data sets expand to more volumes from were cataloged at the time of the original execution, any attempt to stack further data sets by using VOL=REF= results in another abend because the reference uses the catalog information from the beginning of the job step. The system catalogs the expanded data sets again at the end of the job step. However, submitting the job a third time should result in the utility executing with the volumes resolved correctly.

  • For uncataloged stacked tape copies, you must include the VOL=SER information of completed copies in the DD statements before restarting. You must also change the NEW disposition to OLD.

This method allows processing to continue with the failed command, minimizing unnecessary processing. This method usually requires manual intervention during restart to modify the JCL.

BMC AMI Recover installation options that affect restart—CHECKPT and CHECKINT

The CHECKPT and CHECKINT  installation options affect the restart processing for BMC AMI Recover jobs. Using these options minimizes the time that is lost redoing work when a job fails and must be restarted.

For more details about CHECKPT and CHECKINT installation options, see BMC AMI Recover installation options.

You can specify the CHECKPT installation option to:

• Take no checkpoints

• Take checkpoints only at the end of each phase

You can override the CHECKPT installation option at run time by specifying the BMC AMI Recover checkpoint override parameter that is described in Checkpoint override parameter (checkPt).

When you request checkpoints, you can prevent unnecessary checkpoints by specifying a value for the checkpoint interval installation option, CHECKINT, which sets the number of minutes elapsed between checkpoints. You can override the CHECKINT installation option at run time by specifying CHECKINT on the OPTIONS command, as described in CHECKINT (integer).

To determine values for either or both of these options, you must balance the cost of taking checkpoints against the time that is lost redoing work when an BMC AMI Recover execution must be restarted.

Restart considerations with INDEXLOG AUTO

When BMC AMI Recover converts a RECOVER INDEX request to a REBUILD INDEX request based on the specification of INDEXLOG AUTO, BMC AMI Recover updates the BMCSYNC table for the index to indicate that it is now a REBUILD INDEX. During restart, BMC AMI Recover uses the BMCSYNC table update to discover that an index recovery was converted.

If a RECOVER INDEX request that uses an image copy fails during execution, the restart logic causes the index to become a candidate for conversion because the original image copy is invalid.

If a RECOVER INDEX DSNUM ALL request for a nonpartitioned index is converted to multiple DSNUM n requests, and if one of the resulting MERGE phases later fails, restart does not allow the request to be converted to a REBUILD INDEX request. If a multi-data-set index is recovered with a DSNUM 0 image copy, the RECOVER INDEX request can convert to a REBUILD INDEX request.