Considerations for using multicluster control data sets

Multicluster control data sets are a related group of VSAM clusters with the following requirements:

The control data sets cannot be empty when you start DFSMShsm. The MCDS must be defined such that the high key for the first cluster must be greater than X'10' || C'MHCR'.
All clusters in a multicluster CDS must be cataloged in the same user catalog.
All clusters must be on DASD of the same category.
Only primary allocation can be specified for the index and data; no secondary allocation can be specified for either the index or data.
The index and data are limited to one volume; they cannot be on different volumes.
For a multicluster BCDS, the RECORDSIZE and data CONTROLINTERVALSIZE for each cluster depend on the maximum number of backup versions you want to be able to keep for any data set.
All key must be contiguous and must start with X'00' and end with X'FF'.
The maximum record size must be the same for all clusters.
The record key must start at 0 and be 44 characters in length.
The CISIZE value must be the same for the data and index components for all clusters.
The CONTROLINTERVALSIZE must be the same for all clusters of a multicluster CDS.
The volume device type must be the same for all clusters. For example. if volumes are defined as shared, then all volumes must be defined as shared.
Only primary allocation can be specified for non-RLS control data sets; no secondary allocation can be specified for non-RLS control data sets.

Message ARC0130I will be issued if the above rules are not followed.

Attention: Message ARC0264A is issued to confirm the new cluster count after the initialization of the first DFSMShsm host when the number of CDS clusters has been update. Verify the new cluster count indicated in the message is the intended number of clusters. If new cluster count is incorrect, enter N and verify the configuration changes to avoid CDS corruption.

New customers who have determined that their control data sets will span volumes when they fully implement DFSMShsm should do one of the following tasks:

Initially define each CDS on a single volume. As the MCDS and BCDS grow, you can reorganize them into multiple clusters in order to span volumes.
Define the control data sets as single cluster and multivolume. To do this, define the control data sets to use extended addressability (EA) and to use any of the following methods to access them: record level sharing (RLS), CDSQ serialization, or CDSR serialization.

For more information about control data set placement in a multiple DFSMShsm-host environment, see CDS considerations in a multiple DFSMShsm host environment.