Scheduler-driven Cloning

Automated Cloning Process
Subsystem cloning is based on volume copies. The first task is to identify the volumes that need to be copied in order to generate the jobs that duplicate the volumes. Then it is necessary to rename the datasets and to catalog them. Afterwards all references to the original system’s datasets within DB2, and, if involved, IMS, must be changed to the names of the datasets of the clone. Only the first part of the dataset names is to be changed, the FLQ or VCAT. So trivial this sounds, but so intricate is the execution of the task. DB2 does not allow it to simply change its catalog. Creating new storagegroups, stopping all table spaces and indexes, treating all table spaces and indexes with the Alter command, re-starting all of them are only a few of the required working steps. Keep in mind that an SAP application system consists of more than 70,000 datasets, and for the described changes the target system must be running. As all changes are done using DB2 mechanisms, and you know just how long this will take and how much JCL and control statements will be involved!

Reasonably, there is no way around using a DB2 subsystem cloning tool. Besides resolving name conflicts on ICF catalog level, such a tool has functions to perform the internal DB2 name changes fast and thoroughly. Furthermore it is preferable to use a solution that makes the necessary updates to the DB2 directory and catalog tablespaces while DB2 (the clone) is still shut down, using specialized programs rather than slower SQL and DB2 utilities. Key is this method does not require a start and stop of the target system, to run the rename process. As soon as the target DB2 subsystem comes up, it is ready to go.

Parallel processing is required to ensure short cloning times. Cloning of large environments may result in a hundred or more jobs, some need to be executed in a certain sequence. It is best practice to pass the job chains generated by the cloning tool to the scheduler. This way system cloning becomes a scheduler-driven, i.e. a fully automated task.