Migrate data from an Oracle database to an Oracle tenant of OceanBase Database

This topic describes how to use OceanBase Migration Service (OMS) to migrate data from an Oracle database to an Oracle tenant of OceanBase Database.

Background

You can create a data migration project in the OMS console to seamlessly migrate the existing business data and incremental data from an Oracle database to an Oracle tenant of OceanBase Database by using the schema migration, full migration, and incremental data synchronization features.

The Oracle database supports the following modes: primary database only, standby database only, and primary/standby databases. The following table describes the data migration operations supported for each mode.

Prerequisites

• You have created a corresponding schema in the destination Oracle tenant of OceanBase Database. OMS allows you to migrate tables and columns. Therefore, you must create a corresponding schema in the destination database before migration.

• You have enabled archivelog for the source Oracle instance and switched the logfile before OMS starts incremental data replication.

• You have installed LogMiner in the source Oracle instance, and LogMiner runs properly.

  LogMiner enables you to obtain data from the archived logs of the Oracle instance.

• You have created a dedicated database user in the source Oracle database and the destination Oracle tenant of OceanBase Database for data migration and granted the corresponding privileges to the users. For more information, see Create a database user.

• You have made sure that the Oracle instance has enabled the database-level or table-level supplemental_log feature.

• If you enable supplemental_log of the primary key and unique key at the database level, when a large number of unnecessary logs are generated by tables that do not need to be synchronized, the pressure on LogMiner Reader to fetch logs and on the Oracle database increases. Therefore, you can enable only the table-level supplemental_log of the primary key and unique key for Oracle databases in the OMS console. However, if you configure the Set ETL Options to filter columns other than the primary key and unique key columns when you create a migration task, enable supplemental_log for the corresponding columns or all columns.

• Clock synchronization (such as the NTP service) is required between an Oracle server and the OMS server to avoid data risks. For an Oracle RAC, clock synchronization is also required between Oracle instances.

Limits

• OMS supports the following Oracle database versions: 10g, 11g, 12c, 18c, and 19c. Version 12c and later provide container databases (CDBs) and pluggable databases (PDBs).

• OMS allows you to migrate only databases and tables whose names are ASCII codes and do not contain special characters such as .|"'`()=;/&\n.

• OMS does not support triggers in the destination database for long-term synchronization between these databases.

• Data type limits

  • Incremental data migration is not supported for a table whose data in all columns is of the following three large object (LOB) types: BLOB, CLOB, and NCLOB.

  • If a table does not have the primary key and contains data of the LOB type, the reverse incremental migration of the table can suffer poor data quality.

  • If the LOB field in the source Oracle database is too large, it cannot be stored in OceanBase Database, causing data synchronization errors.

• If a new table without the primary key is added in the source Oracle database during the incremental synchronization, OMS does not automatically delete the hidden columns and the unique index added to the table in the destination Oracle tenant of OceanBase Database. You need to manually delete them before you start a reverse migration task.

  To confirm the tables without the primary key that are added during the incremental synchronization, view the manual_table.log file in the logs/msg/ directory.

• Character encoding and reverse synchronization are limited:

  If the source and destination databases use different character sets, a field length extension policy will be provided during schema migration. For example, the field length is extended by 1.5 times, and the length unit is changed from BYTE to CHAR.

  This ensures that data encoded by using different character sets can be migrated from the source database to the destination database. However, after cutover, data may fail to be written back to the source database during reverse incremental data migration because of an extra long data length.

• The data source identifiers and user accounts must be globally unique in OMS.

• OMS can parse up to 5 TB of incremental logs of Oracle databases per day.

• When Oracle Database is of a version earlier than 11g, the database object name cannot exceed 30 bytes in length. In the reverse incremental synchronization step, you cannot create database objects with an object name longer than 30 bytes in the Oracle tenant of OceanBase Database.

• OMS supports full synchronization of tables whose partition fields are not within the primary keys, but does not support DDL operations that create tables.

Considerations

• When the Oracle database is in standby database only or primary/standby databases mode , if the number of instances that run on the primary Oracle database differs from that on the standby database, incremental logs of some instances may not be pulled. You need to manually set the parameters of Oracle Store to specify the instances for which incremental logs are to be pulled from the standby database. The procedure is as follows:

  1. Stop Oracle Store as soon as it starts.

  2. On the Update Configuration page of the Store component, add the deliver2store.logminer.instance_threads parameter and specify the instances for which logs are to be pulled.

     Separate multiple threads with a vertical bar (|), for example, 1|2|3. For more information about how to update a store component, see Update a store component.

  3. After you set the parameters, restart Oracle Store.

  4. Wait for five minutes, and then run the grep 'log entries from' connector/connector.log command to check the instance logs pulled. The thread field indicates the instances for which the logs are pulled.

• If you need to migrate incremental data from an Oracle database, we recommend that you restrict the size of a single archive file in the Oracle database within 2 GB. An excessively large archive file may incur the following risks:

  • The log pulling time increases not in proportion to the size of a single archive file, but much more sharply.

  • When the Oracle database in standby database only or primary/standby databases mode , the incremental data is pulled from the standby database. In this case, only archive files can be pulled. An archive file is pulled after it is generated. A larger archive file means a longer delay before the archive file is processed, and a longer time for processing the archive file.

  • In addition, a larger size of a single archive file means larger memory required by Oracle Store under the same data pulling concurrency.

• The archive files must be stored for more than two days in the Oracle database. Otherwise, in the case of a sharp increase in the number of archived files or an exception in Oracle Store, restoration may fail due to the lack of required archive files.

• OMS allows you to migrate data from the source Oracle instance that uses character sets including AL32UTF8, AL16UTF16, ZHS16GBK, and GB18030. If the character set used by the source database is UTF-8, we recommend that you use UTF-8 or a greater character set for the destination database.

• If you select a migration mode that supports incremental synchronization and reverse incremental migration, and an exception occurs when OMS pulls the incremental data from a standby Oracle database, you can run the ALTER SYSTEM SWITCH LOGFILE command in the primary database to handle the exception.

• When you migrate a table without a primary key from an Oracle database to an Oracle tenant of OceanBase Database, do not perform any operations that may change the ROWID, such as data import and export, Alter Table, FlashBack Table, and partition splitting or compaction.

• If the clocks are out of synchronization between the nodes or between the client and the server, a negative delay may occur in incremental synchronization or reverse incremental synchronization.

• Daylight Savings Time (DST) was once adopted in China, so a one-hour time difference between the source and the destination is expected for the data of the TIMESTAMP(6) WITH TIME ZONE type that was generated during the following periods: the DST periods from 1986 to 1991, and April 10 to 17, 1988.

• In a project of reverse incremental migration from an Oracle database to an Oracle tenant of OceanBase Database, when the Oracle tenant is of a version earlier than V3.2x and has a multi-partition table with global unique indexes, if you update the value of a partitioning key of the table, data may be lost during migration.

• When the destination Oracle tenant of OceanBase Database is of a version earlier than V2.2.70, foreign keys, checks, and other objects added during the switchover may not be supported.

• If you change the unique index of the destination, you must restart the incremental synchronization. Otherwise, the data may be inconsistent.

• If forward switchover is not started in a data migration project, delete the unique indexes and pseudocolumns from the destination database. If you do not delete the unique indexes and pseudocolumns, data cannot be written, and pseudocolumns will be generated again when data is imported to the downstream system, causing conflicts with the pseudocolumns in the source database. When a data migration project is in the forward switchover step, OMS will automatically drop the hidden columns and unique indexes based on the type of the data migration project. For more information, see Schema migration mechanisms.

• If you select migration objects based on matching rules, we recommend that you create standard CTAS statements and set object exclusion rules, such as a.ctas*, to avoid interrupting the data migration project.

• If you use Oracle Database 12c or later and have configured Incremental Migration, perform the following operations to migrate incremental database objects whose names are more than 30 bytes in length.

  1. Execute the following statement as the SYS user to modify parameters of the Oracle database: ALTER SYSTEM SET ENABLE_GOLDENGATE_REPLICATION=true SCOPE=BOTH;.

     Notice:

     For Oracle Real Application Clusters (RAC), you need to modify the parameters for all Oracle instances. You do not need to restart the Oracle database.

  2. Execute the ALTER SYSTEM SWITCH LOGFILE; statement three times for each Oracle instance.

  3. After 10 minutes, re-create a data migration project.

  4. Set deliver2store.logminer.need_check_object_length to false for Oracle Store.

• In a multi-table aggregation scenario:

  • We recommend that you map objects in the source and destination databases by importing objects and configuring matching rules.

  • We recommend that you manually create schemas in the destination database. If you use OMS to create schemas, skip failed objects in the schema migration step.

• Check the objects in the recycle bin of the Oracle database. If the recycle bin contains more than 100 objects, internal table queries may time out. You must clear the objects in the recycle bin.

  • Query whether the recycle bin is enabled.

  SELECT Value FROM V$parameter WHERE Name = 'recyclebin';
  

  • Query the number of objects in the recycle bin.

  SELECT count(*) FROM RECYCLEBIN;
  

• If you skip the "Check the ROW_MOVEMENT parameter of the database" check item of the source primary database during the migration, data inconsistency may occur when you synchronize tables whose ROW_MOVEMENT is enable.

Data type mapping

Conversion of Oracle table partitions

When OMS is used to migrate data from an Oracle database, the system automatically converts your business SQL statements. However, the conversion performed in OceanBase Database V2.2.30 is different from that in OceanBase Database V2.2.50.

Check and modify the system configurations of the Oracle instance

Perform the following operations:

1. Enable archivelog for the source Oracle database.

2. Enable supplemental_log in the source Oracle database.

3. Set the system parameters of the Oracle database.

4. Restart the instance and perform the archivelog switchover three times.

Enable archivelog for the source Oracle database

select log_mode from v$database;

The value of the log_mode field must be archivelog. Otherwise, perform the following steps to change it:

1. Run the following commands to enable archivelog.

   shutdown immediate;
   startup mount;
   alter database archivelog;
   alter database open;
   

2. Run the following command to view the path and quota of archived logs.

   View the path and quota of the recovery file. We recommend that you set the db_recovery_file_dest_size parameter to a relatively large value. After you enable archivelog, you need to regularly clear the archived logs by using RMAN or other methods.

   show parameter db_recovery_file_dest;
   

3. Change the quota of archived logs as needed.

   alter system set db_recovery_file_dest_size=50G scope=both;
   

Enable supplemental_log in the source Oracle database

LogMiner Reader allows you to enable only table-level supplemental_log in an Oracle database. If you create new tables in the Oracle instance before the migration, you must enable supplemental_log for the primary key and unique key before you perform DML operations. Otherwise, OMS returns an exception of incomplete logs.

If the indexes are inconsistent between the source and destination databases, the ETL does not meet the expectation, or the migration performance of partitioned tables deteriorates, you need to add the following supplemental_logs:

• Add the database-level or table-level supplemental_log_data_pk and supplemental_log_data_ui.

• Add columns to the supplemental_logs.

  • Add all columns involved by the primary keys or unique keys in the source and destination databases to resolve the problem of index inconsistency between the source and destination databases.

  • If an ETL exists, add the ETL column to resolve the problem that the ETL does not meet the expectation.

  • If the destination table is a partitioned table, add a partition column to resolve the problem that the write performance deteriorates because partition pruning cannot be performed.

  You can execute the following statement to check the addition result.

  select log_group_type from all_log_groups where owner = 'Database' and table_name = 'Table';
  

  If the check result includes ALL COLUMN LOGGING, the check is passed. Otherwise, check whether the ALL_LOG_GROUP_COLUMNS table contains all preceding columns.

  Sample statement for adding columns to supplemental_logs:

  alter table <table_name> add supplemental log group <table_name_group> (c1, c2) always;
  

The following table describes the possible risks and solutions when you perform DDL operations in a running data migration project.

LogMiner Reader uses one of the following two methods to check whether supplemental_log is enabled. If not, LogMiner Reader exits.

• Enable supplemental_log_data_pk and supplemental_log_data_ui at the database level.

  Run the following commands to check whether the supplemental log feature is enabled. If the returned values are both YES, supplemental_log is enabled.

  select supplemental_log_data_pk, supplemental_log_data_ui from v$database;
  

  Otherwise, perform the following steps:

  1. Execute the following statement to enable the supplemental log feature.

     alter database add supplemental log data(primary key, unique) columns;
     

  2. Perform switchover to the ARCHIVELOG mode three times. For an Oracle RAC, perform switchover for the instances alternately.

     alter system switch logfile;
     

     The reason for performing the archivelog switchover three times:

     When Oracle Store locates the start time for pulling log files, the database is rolled back by 0 to 2 archive logs based on the specified timestamp. Therefore, after you enable the supplemental log feature, you need to perform switchover to the ARCHIVELOG mode three times to prevent Store from pulling the logs that are generated before the specified timestamp. Otherwise, Store exits unexpectedly.

     The reason for alternately performing the archivelog switchover among multiple instances in an RAC system:

     In an Oracle RAC system, if you perform the archivelog switchover multiple times on one instance, when you perform the archivelog switchover on the next instance, the latter instance may pull the logs that are generated before supplemental_log is enabled.

• Enable supplemental_log_data_pk and supplemental_log_data_ui at the table level.

  1. Execute the following statement to confirm whether supplemental_log_data_min is enabled at the database level.

     select supplemental_log_data_min from v$database;
     

     If the returned value is YES or IMPLICIT, supplemental_log is enabled.

  2. Execute the following statement to check whether the table-level supplemental_log is enabled for the tables to be synchronized.

     select log_group_type from all_log_groups where owner = 'xxx' and table_name = 'yyy';
     

     Each type of supplemental_log returns one row. The results must contain ALL COLUMN LOGGING or both PRIMARY KEY LOGGING and UNIQUE KEY LOGGING.

     If the table-level supplemental_log is not enabled, execute the following statement.

     alter table table_name add supplemental log data(primary key, unique) columns;
     

  3. Perform switchover to the ARCHIVELOG mode three times. For an Oracle RAC, perform switchover for the instances alternately.

     alter system switch logfile;
     

(Optional) Set the system parameters of the Oracle database

We recommend that you set the _log_parallelism_max parameter of the Oracle database to 1. The default value is 2.

You can use one of the following two methods to query the value of the _log_parallelism_max parameter.

• Method 1

  SELECT NAM.KSPPINM,VAL.KSPPSTVL,NAM.KSPPDESC FROM SYS.X$KSPPI NAM,SYS.X$KSPPSV VAL WHERE NAM.INDX= VAL.INDX AND NAM.KSPPINM LIKE '_%' AND UPPER(NAM.KSPPINM) LIKE '%LOG_PARALLEL%';
  

• Method 2

  select value from v$parameter where name = '_log_parallelism_max';
  

Execute one of the following statements to modify the value of the _log_parallelism_max parameter.

• Oracle RAC

  alter system set "_log_parallelism_max"=1 sid='*' scope=spfile;
  

• Non-Oracle RAC

  alter system set "_log_parallelism_max"=1 scope=spfile;
  

When you modify the value of the _log_parallelism_max parameter in Oracle Database 10g, if the error message write to SPFILE requested but no SPFILE specified at startup is returned, perform the following operations:

create spfile from pfile;
shutdown immediate;
startup;
show parameter spfile;

Restart the instance and perform log switchover

After completing the preceding operations, restart the instance and perform log switchover three times.

Create a data migration project

1. Create a migration project.

   1. Log on to the OMS console.

   2. In the left-side navigation pane, click Data Migration.

   3. On the Data Migration page, click Create Migration Project in the upper-right corner.

2. On the Select Source and Destination page, specify the following parameters.

   Parameter	Description
   Migration Project Name	We recommend that you set it to a combination of digits and letters. It must not contain any spaces and cannot exceed 64 characters in length.
   Tag	Click the field and select a target tag from the drop-down list. You can click Manage Tags to create, modify, and delete tags. For more information, see Use tags to manage data migration projects.
   Source	If you have created an Oracle data source, select it from the drop-down list. If you have not created an Oracle data source, click Create Data Source in the drop-down list and create one in the dialog box that appears on the right. For more information about the parameters, see Create an Oracle data source.
   Destination	If you have created a data source for the Oracle tenant of OceanBase Database, select it from the drop-down list. If you have not created a data source for the Oracle tenant, click Create Data Source in the drop-down list and create one in the dialog box that appears on the right. For more information, see Create a physical data source of OceanBase Database. Notice The destination cannot be an Oracle tenant of OceanBase Database V4.0.0.

3. Click Next. On the Select Migration Type page, specify the following parameters.

   Options available for Migration Type include Schema Migration, Full Migration, Incremental Synchronization, Full Verification, and Reverse Incremental Migration.

   Migration type	Description
   Schema Migration	The definitions of data objects, such as tables, indexes, constraints, comments, and views, are migrated from the source database to the destination database. Temporary tables are automatically filtered out.
   Full Migration	If you select Full Migration, we recommend that you use the GATHER_SCHEMA_STATS or GATHER_TABLE_STATS statement to collect the statistics of the Oracle database before data migration.
   Incremental Synchronization	Options for Incremental Synchronization are DML Synchronization and DDL Synchronization. You can select the options as needed. For more information about DDL synchronization, see Synchronize DDL operations from an Oracle database to an Oracle tenant of OceanBase Database. Incremental Synchronization has the following limits: For Oracle 12c or later, when you add or change a column, the table name and column name cannot exceed 30 bytes in length. If you want the database to support table names and column names of more than 30 bytes in length, specify the ENABLE_GOLDENGATE_REPLICATION parameter as the SYS user in the Oracle database, and set deliver2store.logminer.need_check_object_length=false for Oracle Store. Set ENABLE_GOLDENGATE_REPLICATION as follows: For a Real Application Cluster (RAC) environment, set this parameter for each node. If the Oracle database is in Active Data Guard (ADG) mode, set this parameter in the ADG source database. ALTER SYSTEM SET ENABLE_GOLDENGATE_REPLICATION=true SCOPE=BOTH; Query ENABLE_GOLDENGATE_REPLICATION as follows. SELECT K.KSPPINM,V.KSPPSTVL FROM SYS.X$KSPPI K,SYS.X$KSPPSV V WHERE K.INDX=V.INDX AND UPPER(K.KSPPINM) = 'ENABLE_GOLDENGATE_REPLICATION'; If you select DDL Synchronization, when you perform a DDL operation that is not supported by OMS in the source database, data migration may fail. If the DDL operation creates a new column, we recommend that you set the attribute of the column to Null. Otherwise, data migration may be interrupted. The source Oracle database does not support incremental synchronization of tables using the empty_clob() function.
   Full Verification	If you select Full Verification, we recommend that you collect the statistics of the Oracle database and the Oracle tenant of OceanBase Database before full verification. For more information about how to collect statistics of an Oracle tenant of OceanBase Database, see Manually collect statistics. If you have selected Incremental Synchronization but did not select all DML statements in the DML Synchronization section, OMS does not support full verification.
   Reverse Incremental Migration	If a table to migrate has no primary key or unique index and a large amount of data in the table is changed, the reverse incremental migration will take a long time. In this case, you can add unique indexes in the source database. You cannot select Reverse Incremental Migration in the following cases: Multi-table aggregation and synchronization is enabled. Multiple schemas are configured in a rule to match one type of objects.

4. (Optional) Click Next. If you have selected Reverse Incremental Migration but the ConfigUrl, username, or password is not configured for the data source of the destination Oracle tenant of OceanBase Database, the More about Data Sources dialog box appears, prompting you to configure related parameters. For more information, see Create a physical data source of OceanBase Database.

   After you configure the parameters, click Test Connectivity. After the test succeeds, click Save.

5. Click Next. On the Select Migration Objects page, select the migration objects and migration scope.

   You can select one of the following two modes to migrate objects: Specify Objects or Match Rules. If you have selected DDL Synchronization, only the Match Rules option is available.

   • Select Specify Objects. Then select the objects to be migrated on the left and click > to add them to the list on the right. You can select tables and views of one or more databases as the migration objects.

     Notice

     • The names of tables to be migrated, as well as the names of columns in the tables, must not contain Chinese characters.

     • If the database or table name contains a double dollar sign ($$), you cannot create the migration project.

     When you migrate data from an Oracle database to an Oracle tenant of OceanBase Database, OMS allows you to import objects through text, rename object names, set row filters, view column information, and remove one or all objects to be migrated.

     Operation	Steps
     Import Objects	In the list on the right of the Specify Migration Scope section, click Import Objects in the upper-right corner. In the dialog box that appears, click OK. Notice This operation will overwrite previous selections. Proceed with caution. In the Import Objects dialog box, import the objects to be migrated. You can import CSV files to rename databases/tables and set row filtering conditions. For more information, see Download and import the settings of migration objects. Click Validate. After the validation succeeds, click OK.
     Rename	OMS allows you to rename the migration objects. For more information, see Rename a database table.
     Settings	OMS allows you to use the WHERE clause to filter rows. For more information, see Use SQL conditions to filter data You can also view column information of the migration object in the View Column section.
     Remove/Remove All	During data mapping, OMS allows you to remove one or more selected objects to be migrated to the destination. To remove a single migration object: In the list on the right of the Specify Migration Scope section, hover the pointer over the target object, and click Remove. To remove all migration objects: In the list on the right of the Specify Migration Scope section, click Remove All in the upper-right corner. In the dialog box that appears, click OK.

     When the source database is an Oracle database, if row filtering is enabled for columns other than the primary key and unique key columns, enable supplemental_log for the corresponding columns or all columns.

     Statement for enabling supplemental_log for the corresponding columns:

     ALTER TABLE table_name ADD SUPPLEMENTAL LOG GROUP log_group_name (column1, column2, column3) ALWAYS;
     

     Statement for enabling supplemental_log for all columns:

     --Enable database-level supplemental_log:
     ALTER DATABASE ADD SUPPLEMENTAL LOG DATA (ALL) COLUMNS;
     --Enable table-level supplemental_log:
     ALTER TABLE table_name ADD SUPPLEMENTAL LOG DATA (ALL) COLUMNS;
     

   • Select Match Rules. For more information, see Configure matching rules for migration objects.

6. Click Next. On the Migration Options page, specify the following parameters.

   Parameter	Description
   Incremental Synchronization Start Timestamp	If you have set the migration type to Full Migration, this parameter is not displayed. If you have selected a migration type other than Full Migration, specify a point in time after which the data is to be synchronized. The default value is the current system time. You can select a point in time or enter a timestamp. Notice You can select the current time or a point in time earlier than the current time. This parameter is closely related to the retention period of archived logs. Generally, you can start data synchronization from the current timestamp.
   Concurrency for Full Migration	The value can be Smooth, Normal, or Fast. The quantity of resources to be consumed by a full data migration task varies based on the migration performance. You can also modify the configurations of the Full-Import component to customize the concurrency. Notice: To enable this feature, select Full Migration on the Select Migration Type page.
   Full Verification Concurrency	The value can be Smooth, Normal, or Fast. Different quantities of resources of the source and destination databases are consumed at different concurrencies. You can also modify the configurations of the Full-Verification component to customize the concurrency.
   Incremental Record Retention Time	The duration that incremental parsed files are cached in OMS. A longer retention period indicates more disk space occupied by the Store component of OMS.
   Whether to Allow Destination Table to Be Not Empty During Full Migration	If destination tables are allowed to be not empty during full migration, full verification is performed in IN mode. Notice: To enable this feature, select Full Migration on the Select Migration Type page.
   Whether to Allow Post-indexing	You can specify whether to allow post-indexing after full migration is completed. Post-indexing can shorten the time of full migration. Notice: To enable this feature, select both Schema Migration and Full Migration on the Select Migration Type page. Only non-unique key indexes can be created after the migration is completed.

7. Click Precheck to start a pre-check on the data migration project.

   During the pre-check, OMS checks the read and write privileges of the database users and the network connections of the databases. The data migration project can be started only after it passes all check items. If an error is returned during the pre-check:

   • You can identify and troubleshoot the problem and then perform the pre-check again.

   • You can click Skip in the Actions column of the pre-check item with the error. A dialog box will be displayed, prompting the impact caused if you skip this error. If you want to continue, click OK in the dialog box.

8. Click Start Project. If you do not need to start the project now, click Save to go to the details page of the data migration project. You can start the project later as needed.

   OMS allows you to modify the migration objects when the data migration project is running. For more information, see View and modify migration objects. After a data migration project is started, the migration objects will be executed based on the selected migration type. For more information, see the "View migration details" section in the View details of a data migration project topic.