OBLOADER & OBDUMPER V4.3.6

OBLOADER & OBDUMPER V4.3.6 is released with enhancements in automatic column mapping, expanded import and export capabilities, improved support for HBase and OBKV scenarios, and optimizations for memory and stability. It also addresses a number of known issues to enhance compatibility, stability, and usability across various migration and import/export scenarios.

New features

• Added support for automatic column mapping in import scenarios where the number of columns in the source file does not match the target table, improving compatibility with non-standard data files.

• Added support for automatic column mapping for files without header column information, matching columns based on order, reducing the need for additional data file preprocessing.

• Extended automatic column mapping to SQL and AVRO format files, enhancing the generality of import capabilities.

• Adjusted the default value of the --parallel parameter from 1 to 20 to optimize task parallel processing capabilities.

• Added the --exclude-object-types option to allow filtering of specified database object types during import and export.

• Added the --column-case-sensitive option to specify column name case sensitivity when exporting CSV files.

• Added the --hbase-cf option to specify the ColumnFamily value when writing to HBase, supporting OBKV-HBase import scenarios.

• Added the --hbase-rowkey option to specify the rowkey column name when writing to HBase, supporting OBKV-HBase import scenarios.

• Added the --hbase-rowkey-reverse option to specify whether to reverse the rowkey when writing to HBase, meeting different OBKV-HBase scenario requirements.

• Added the --skip-header option in POS format files to support skipping the first row of data.

• In CSV format, added support for exporting column names without specifying --skip-header when using the --retain-empty-files option.

• Extended the --ignore-escape option from being applicable only to import to a general option for both import and export, allowing users to choose whether to escape as needed.

• In table definition export, added support for controlling the column value generation strategy through --sequence-policy, where preserve means exporting based on the current sequence progress, and restart means exporting based on the initial value at table creation.

• Added a buffer size limit mechanism to control memory usage and reduce the risk of OOM during memory expansion and data import.

Bug fixes

• Fixed the issue where the program would hang if the tenant name was not specified during direct load.

• Fixed the issue where number type data exported from an Oracle tenant would be converted to scientific notation.

• Fixed the issue where the -D option was required in Oracle compatibility mode, reverting it to an optional option.

• Fixed the issue where uppercase primary key matching column positions could not be correctly retrieved in MySQL mode, causing an error.

• Fixed the issue where the error "The object type name is null" would occur during OSS data import.

• Fixed the issue where using a non-existent directory would still result in the error "The output directory is not empty" during export.

• Fixed the issue where the --replace-object and --drop-object options would fail when importing data containing special characters like timeout.

• Fixed the issue where the error would occur directly without specifying the --escape-character option when exporting with the --column-quote-mode=none option, changing it to a warning instead.

• Fixed the issue where the startup script would report the error VM option 'UseG1GC' is experimental and must be enabled via -XX:+UnlockExperimentalVMOptions for GC_OPTS.

Dependency library upgrades

• Upgraded the version of the commons-compress component from 1.22 to 1.26.0.

• Upgraded the version of the oceanbase-client component from 2.4.14 to 2.4.17.

• Upgraded the version of the hadoop component from 3.3.6 to 3.4.0.