Restrictions for a partitioned table

When you use partitioned tables, be aware of these restrictions.

Referential constraints are allowed for a partitioned table, however, the parent key index must be a nonpartitioned index.
If a primary key constraint for a partitioned table is added and then dropped, the primary key index is also dropped.
If a primary key constraint is added to a partitioned table, and then removed by the user, the user is not allowed to keep the table keyed.
If an existing nonpartitioned table does not have a primary key constraint, but the table is keyed, the keys are removed when the table is changed to a partitioned table.
Db2 Multisystem files (distributed tables) are already partitioned across multiple systems and cannot be partitioned across multiple members on a single system.
An update to the partitioning key that attempts to move a row to a different partition is allowed if the partitioned table is journaled.
The number of partitioning keys is restricted to 120.
All SQL relative record processing is handled as it is for Db2 Multisystem support. The relative record number is determined in each individual partition, not the table as a whole. For example, reading to record 27 means that you read to record 27 in each partition. Each partition can contain its own record 27, none of which is the same.
There are some restrictions on the data type of a partition key column. For range partitioning, the data type of a column used to partition a table cannot be BLOB, CLOB, DBCLOB, DATALINK, floating-point type, or a distinct type based on the these types. For hash partitioning, the data type of the column used as part of the partition key cannot be LOB, DATE, TIME, TIMESTAMP, floating-point type, or a distinct type based on one of these.
Applications using the following CL commands must be changed to use Member *ALL to process all partitions of a partitioned table: