There are some extra considerations you need to take into account if you are implementing a Mendix app using a DB2 database. In addition, the behavior of Mendix using a DB2 database has some minor differences when compared with using a PostgreSQL database.
These considerations and differences are documented below.
2 Page Size of the Table Space
For Mendix to run on DB2, it is very important that the user table space has a page size of at least 8K (but preferably 32K). This is because Mendix uses national character strings (NVARCHAR or VARCHAR with string units CODEUNIT32). This data type consumes more space than an octets-based VARCHAR. For the system administration, Mendix always creates some tables with indexes, which require at least a table space page size of 8K.
Exception with SQL Code
If the index is too big for the page size, DB2 will throw an exception:
com.ibm.db2.jcc.am.SqlException: DB2 SQL Error: SQLCODE=-614, SQLSTATE=54008, SQLERRMC=some_index_name
For user-created indexes, if the combined length of the specified columns in the index is greater than the max key length, you should also increase the page size of the table space.
For more detailed information, see SQL0614N – The index or index extension index-name cannot be created or altered because the combined length of the specified columns is too long in the SQL messages section of the IBM Knowledge Center.
3 Transaction Log Size
Exception with SQL Code
If the transaction log space is depleted or there is a temporary increase in the number of active transactions, DB2 will throw this exception:
com.ibm.db2.jcc.am.SqlException: DB2 SQL Error: SQLCODE=-964, SQLSTATE=57011, SQLERRMC=null.
In this case, the size of LOGPRIMARY must be increased.
For more detailed information, see DB2 SQL error: SQLCODE: -964, SQLSTATE: 57011, SQLERRMC: null on the IBM Support pages and SQL0964C – The transaction log for the database is full in the SQL messages section of the IBM Knowledge Center.
5.1 String Comparison Is Case-Sensitive
Sorting on string column values is case-sensitive in DB2. To mitigate this, IBM has introduced locale-aware Unicode collations in DB2 9.5 fixpack 1. These collations can be tailored to ignore case and/or accents.
For more detailed information, see the sidenote in Making DB2 case-insensitive in IBM Developer Works.
5.2 Sorting on Very Long Strings
It is not possible to sort on unlimited strings or strings with a specified length greater than 8192 characters. This is because such long or unlimited strings are implemented with the data type NCLOB. DB2 does not allow sorting on columns with this data type. Technically, it is possible to cast this type during the execution of the query to a normal VARCHAR type and sort on this, but this increases the execution time. The question is if it really is user-friendly to show such long strings in a data grid. Consider decreasing the length of the string attribute or removing it from data grids.
5.3 ORDER BY a Correlated Scalar Fullselect or a Function with an External Action
According to the order-by-clause documentation in the IBM DB2 SQL reference, DB2 does not support ORDER BY a correlated scalar fullselect (SQLSTATE 42703) or a function with an external action (SQLSTATE 42845).
Taking this limitation into account, ordering by the associated attribute is not supported when a Mendix application is backed by DB2. Therefore, any associated attribute that is used for ordering is filtered out from the query and the result set is returned as if ordering by the associated attribute had not been presented in the query.
5.4 Non-Blocking Read-Isolated Streaming with OData
According to the Isolation levels documentation in IBM DB2 Application design, non-blocking read-isolated queries are not supported by DB2. The default behavior of DB2 is that when one user is retrieving rows from a table and another user is making modifications on the same table at the same time, then those modifications will show up in the data in the retrieve query (which means database reads are not isolated). Configuring a stricter transaction isolation level to prevent this behavior puts locks on those same rows (which means concurrent database actions are blocking).
Taking this limitation into account, preventing concurrent row modifications from being included in the result set of a data retrieve action is not supported when a Mendix application is using DB2 as a streaming OData datasource.
5.5 Select DISTINCT attribute for Very Long Strings
Selecting DISTINCT attributes of type String with size > 8168 characters is not supported by Mendix due to a known DB2 limitation of selecting DISTINCT columns with a CLOB data type. When you run into this limitation, you may encounter an exception in the logs with a message like this:
DB2 SQL Error: SQLCODE=-727, SQLSTATE=56098, SQLERRMC=2;-134;42907