IBM DB2 Analytics Accelerator for z/OS Version 5.1.0 maintenance level PTF-3: Release Notes

Please find information about new functions here.

1. Snippet-Processing Units (SPUs) running out of resources
   Affected users

   All users of IBM DB2 Analytics Accelerator for z/OS

   Problem description

   Accelerator goes offline

   Problem summary

   This problem occurs if queries that are routed to an accelerator contain numerous recursive REPLACE() calls.

   Problem conclusion

   The problem has been fixed.
2. Accelerator-shadow tables become inconsistent with empty tables in DB2 for z/OS
   Affected users

   All users of IBM DB2 Analytics Accelerator for z/OS

   Problem description

   An affected accelerator-shadow table contains more rows than the DB2 table.

   Problem summary

   This problem occurs during a full table reload, when you reload a segmented table or all partitions of a partitioned table.

   1. A DB2 for z/OS table contains data.
   2. A corresponding accelerator-shadow table exists and you have loaded it.
   3. All data is deleted from the table in DB2 for z/OS.
   4. The now empty table is reloaded on the accelerator. At this point, the DB2 for z/OS is empty, but the accelerator-shadow table might still contain the data loaded in step 2.

   Problem conclusion

   This PTF fixes the problem described in the following tech note: Data inconsistencies between accelerator-shadow tables and empty DB2 for z/OS tables

For recommendations and considerations concerning migration to IBM DB2 Analytics Accelerator for z/OS Version 5.1.0 read here.

1. IBM Netezza Analytics
   For known issues with the IBM Netezza Analytics stored procedure package in IBM DB2 Analytics Accelerator for z/OS Version 5.1.0, see:

   Known issues with IBM Netezza Analytics for System z for IBM DB2 Analytics Accelerator for z/OS Version 5.1.0

2. Encryption of Data in Motion
   For known issues with the network traffic encryption feature (encryption of data in motion) in IBM DB2 Analytics Accelerator for z/OS Version 5.1.0, see:

   Known Issues with Encryption of Data in Motion in IBM DB2 Analytics Accelerator for z/OS Version 5.1.0

3. Transfer of software update packages from IBM DB2 Analytics Accelerator Studio not possible
   Summary

   IBM Analytics Accelerator Studio does not allow users to transfer packages. A message hints at the fact that a newer version is already installed on the server. If the locale in the IBM Analytics Accelerator Data Studio is set to a value different from English or German, the transfer of software update packages might be blocked in rare cases.

   Solution or workaround

   The problem has been fixed with the stored procedures of version 5.1.0 PTF-1. If you run the software transfer with a previous version of the stored procedures, use the following workaround:

   1. Use US English as the locale for IBM DB2 Analytics Accelerator Studio.
   2. On a Windows system, add '-nl en_US' (without the quotes) to the start command that you find in the properties of the shortcut that launches the application ("Target" field on "Shortcut" tab).

      On a Linux system, open the eclipse.ini file and add the following line:

      -Duser.language=en -Duser.country=US

   3. Save and close the file. Then start the application.
4. If the system utilization is very high, the accelerator might restart automatically to recover from the overload situation
   Summary

   The accelerator goes online again after the recovery. Processes that were ended prematurely by the restart, such as load jobs or queries, must be restarted.

   Solution or workaround

   1. Check the system utilization.
   2. Adjust the system utilization as needed.
   3. Retry failed operations and continue.
5. Load operations stall if the system utilization is very high
   Summary

   Occasionally, load jobs might end abnormally if the system utilization is very high.

   Solution or workaround

   Cancel and restart the load operation.
6. The execution of an INSERT INTO ... SELECT ... FROM statement in an in-database transformation job might fail
   Summary

   In rare situations, INSERT INTO ... SELECT ... FROM statements in in-database transformation jobs can fail if the system utilization is very high and many processes are run in parallel.

   Solution or workaround

   This is an indication that the system limits were reached. Check the system utilization and adjust it to the capacity of the system.
7. Incremental update processing might stop and restart if the system utilization is very high (many parallel load and query operations). This is indicated by according console messages
   Summary

   If the system utilization is very high, with many processes running in parallel, the processing of incremental updates might stop. Processing resumes after some time and stalled operations are recovered automatically. If "continuous incremental updates" are enabled, tables that are being loaded at the time of the interruption go into the "suspended" state. You must reload such tables to get them back into the incremental update processing queue.

   Solution or workaround

   This is an indication that the system limits were reached. Check the system utilization and adjust it to the capacity of the system.
8. High system utilization causes slower performance of IBM DB2 Analytics Accelerator tasks
   Summary

   Maintenance tasks might take longer than expected if the system utilization is high.

   Solution or workaround

   If the problem persists, restart the Netezza Platform Software (NPS) from the IBM DB2 Analytics Accelerator Console.
9. Queries might fail if system utilization is high
   Summary

   A high system utilization might cause queries to fail occasionally with errors, such as ‘NZ IO error‘ or ‘SQL code = -904 00E7000E Bad data’ .

   Solution or workaround

   The system continues to operate. Subsequent queries succeed.
10. A reload of a table set with many tables fails
    Summary

    A reload of a table set with thousands of tables might fail sporadically.

    Solution or workaround

    Split the table sets to be reloaded into smaller units and retry the operation.
11. Load operation aborts with SIGABEND if the system utilization is very high
    Summary

    Occasionally, a load operation might end abnormally if the system utilization is very high.

    Solution or workaround

    Restart the load operation.
12. Misleading warnings of type 'DSNX881I  -DI11 2001 W [...] SQL statement with task ID xxxxxxxx is running for more than xxxxxx seconds'
    Summary

    Under rare circumstances, a query might be continually reported as long-running, even though it was ended by a network outage or by a connection closure through the client application. These messages do not indicate that query processing is impacted in any way. The system stays operational and subsequent queries will run as expected.

    Solution or workaround

    The continual reporting of long-running tasks in DSNX88I messages stops after a restart of the accelerator. You can restart an accelerator from the IBM DB2 Analytics Accelerator Console.
13. Insertion of integer and decimal values causes failure of IDT jobs
    Summary

    The insertion of integer and decimal values by an INSERT INTO ... VALUES statement might cause a failure of in-database transformation (IDT) jobs.

    For example, the following statement failed during tests:
    CREATE TABLE T1 (CDEC51 DECIMAL(5,1)) IN ACCELERATOR XYZ;
    INSERT INTO T1(CDEC51) VALUES(77247.04-69993);

    Symptoms: SQL codes like the following are returned:

    SQLCODE = -802, SQLSTATE = 22003, DATA EXCEPTION FROM IBM DB2 Analytics Accelerator TOKENS OVERFLOW ACCELERATOR NUMERIC DATA...

    Solution or workaround

    Use only decimal values instead of a mixture of integer and decimal values. The following statement, for example, will be successful:

    INSERT INTO T1(CDEC51) VALUES(77247.04-69993.0)

14. Tables cannot be added if their names contain special characters, such as tabs and blanks
    Summary

    Tables cannot be added if their names contain special characters, such as tabs and blanks.

    Solution or workaround

    Fix will be delivered in a later release.
15. Large number of accelerator-shadow tables might extend time needed to activate incremental updates
    Summary

    If there is a large number of tables on an accelerator, that is, more than 30000 the activation of incremental updates for these tables might take longer. In this context, it does not matter whether the tables have already been enabled. The extra time needed is relevant if the number of table activations in one request exceeds 10 because this will lead to a premature end of the operation, which in turn leads to an automatic disablement of incremental updates for those tables. You must re-enable and reload such tables.

    Solution or workaround

    Reduce the number of tables that are activated in a single batch job to less than ten (<10).
16. If the system utilization is high, load operations might fail with a message "Failed to start child transaction in session " -LocksharedObjectWait: LockAcquire failed".
    Summary

    Under a high system load, load operations might fail with a message "Failed to start child transaction in session " -LocksharedObjectWait: LockAcquire failed".

    Solution or workaround

    Retry the load operation.

• Use the sample code (for a stored procedure) in the following technote:

  How to store information about accelerated tables in a DB2 for z/OS table

  to retrieve information about your accelerator-shadow tables. This information will be stored in a temporary DB2 table.

• Run a simple SQL query against the replicationStatus and the lastLoadTimestamp columns of this temporary table to identify tables that must be reloaded. Follow the diagram at the end of this note to dertermine whether a table must be reloaded.

• Immediately after the upgrade to V5.1.0 PTF-3, an automatic conversion of data structures takes place.
• Depending on the total amount of data to be converted, this might take more than 1 hour, during which the Netezza Platform Software (NPS) will be in the "Initializing" state. That is, the NPS cannot process requests.