Backup/Archive › VSAM Considerations › Processing Clusters that have Non-Standard Formats

Processing Clusters that have Non-Standard Formats

Certain program products use VSAM but have a non-standard internal format in the VSAM control interval. Among those known to have such a problem are IBM's DB2 database management system, IBM's linear data sets (LDS) and certain data sets created by MSA products. CA Disk, must process these data sets in a special manner.

The typical control interval in the data component of a VSAM cluster looks something like this:

• Record n = data record (actual user data)
• Free space = space that is unused in the control interval
• RDF = Record Descriptor Field. There can be one for each data record in the control interval, but this is not always the case.
• CIDF = Control Interval Descriptor Field. This field normally describes the amount of space being used by the data records portion, as well as the number of free space bytes remaining.

When CA Disk processes a VSAM data set, it can process in one of two modes: logical record or with control interval access. With logical record access, CA Disk issues requests for individual data records and will never see the RDF and CIDF fields. With control interval access, CA Disk is returned the entire control interval with each read request; it must deblock the control interval using the CIDF and RDF control information.

The problem with these non-standard data sets is that they do not maintain the CIDF and/or RDF records, or they use their own techniques in managing the internal data. When CA Disk (or IDCAMS, for that matter!) tries to process these records using logical record techniques, unpredictable results will occur. There is no way of knowing that the CIDF and RDF are not present, or that they have assumed a different meaning from what the initial design of VSAM intended.

The way to avoid these problems is to read the data set with control interval access (as CA Disk has been able to do for some time) and then archive the entire control interval as a single record. In this way CA Disk can ignore what should be the CIDF and RDF records; it does not attempt to interpret the data inside the control interval. The only disadvantage to this technique is that at restore time the control interval size cannot change. This means that CA Disk must put the cluster back exactly the way it archived it.

By default, CA Disk always writes archived data sets in logical record format. CA Disk automatically detects data sets that do not observe logical record definitions, such as DB2 data sets, Oracle data sets and some data sets created by MSA products. When one of these data sets is found, CA Disk automatically shifts to control interval image copy format when archiving the data set. If you have VSAM data sets in non-standard formats that CA Disk cannot detect, you have two options for processing these data sets. You can code sysparm VSARCFMT with a value of C, which causes all VSAM ESDS data sets to be processed in control interval image copy format, or you can code your own user exit called VSACESEX to selectively determine which mode of archiving to use for each data set.

At this time, entry-sequenced data sets (ESDS) are the only type supported by control interval image copy format. KSDS and RRDS clusters are always archived in logical record format.