SQL/MP Installation and Management Guide

Reorganizing Tables and Maintaining Data
HP NonStop SQL/MP Installation and Management Guide—523353-004
8-16
Appending Data to Tables or Partitions
•
Three APPEND parameters control block slack:
°
DSLACK controls the amount of free space in a table’s data blocks.
°
ISLACK controls the amount of free space in a table’s index blocks.
°
SLACK controls the amount of free space in both data and index blocks.
•
You cannot use APPEND to append data to an entire audited table without
resetting the AUDIT attribute, but you can use APPEND to add data to a single
partition of an audited, key-sequenced table.
•
Applications do not have access to a table being modified by APPEND. If you use
the APPEND PARTONLY option to modify a partition, the affected partition is not
accessible to applications while APPEND is in progress; other partitions remain
available.
•
If an error occurs during an append operation, and APPEND is able to terminate
gracefully, no new data is added to the target table. To determine if an append
operation succeeded, check the SQLCI listing to see if error messages occurred.
•
If a processor failure, process failure, BREAK command, or another event
interrupts an append operation, and APPEND cannot terminate gracefully, the
target table remains inaccessible to applications. You must use the
APPENDRESTART or APPENDCANCEL command to restore the target table to its
original state and either complete the append operation (with APPENDRESTART)
or cancel the operation (with APPENDCANCEL).
For more information about the APPENDRESTART and APPENDCANCEL
commands, see the SQL/MP Reference Manual.
For information about loading tables, see Guidelines for Loading Tables on page 8-8.
Appending Data to Multiple Partitions in Parallel
The PARTONLY option lets you append data to partitioned tables in parallel. This
strategy can improve append performance if table partitions are distributed across
disks, processors, and I/O channels. These steps describe possible strategies for
appending data to partitions in parallel:
1. Start an SQLCI process for each partition. One way to do this is to start each
SQLCI process in the processor associated with the target partition. Another way is
to start SQLCI processes in the processors associated with the data sources for
the APPEND command.
2. Issue one APPEND...PARTONLY request for each SQLCI process (and thus each
corresponding partition).