Help Center/ DataArts Studio/ FAQs/ DataArts Migration/ How Do I Select Distribution Columns When Using CDM to Migrate Data to GaussDB(DWS)?
Updated on 2024-08-30 GMT+08:00
View PDF
Share

How Do I Select Distribution Columns When Using CDM to Migrate Data to GaussDB(DWS)?

Symptom

How do I select distribution columns when using CDM to migrate data to GaussDB(DWS)?

Solution

When using CDM to migrate data to DWS or FusionInsight LibrA and create a table on DWS, select the distribution columns on the Map Field tab page.

Figure 1 Selecting distribution columns
Selecting the distribution column is very important for the running of DWS/FusionInsight LibrA. When migrating data to DWS/FusionInsight LibrA, you are advised to specify the distribution column according to the following principles:
  1. Use the primary key as the distribution column.
  2. If multiple data segments are combined as primary keys, specify all primary keys as the distribution column.
  3. In the scenario where no primary key is available, if no distribution column is selected, DWS uses the first column as the distribution column by default. As a result, data skew risks exist.

Therefore, when a single table or entire database is imported to DWS/FusionInsight LibrA, you are advised to manually select a distribution column; otherwise, CDM automatically selects one. For more information about distribution columns, see GaussDB(DWS).

If the DWS primary key or table contains only one field, the field type must be a common character string, value, or date. When data is migrated from another database to DWS, if automatic table creation is selected, the primary key must be of the following types. If no primary key is set, at least one of the following fields must be set. Otherwise, the table cannot be created and the CDM job fails.

  • INTEGER TYPES: TINYINT, SMALLINT, INT, BIGINT, NUMERIC/DECIMAL
  • CHARACTER TYPES: CHAR, BPCHAR, VARCHAR, VARCHAR2, NVARCHAR2, TEXT
  • DATA/TIME TYPES: DATE, TIME, TIMETZ, TIMESTAMP, TIMESTAMPTZ, INTERVAL, SMALLDATETIME
Parent topic: DataArts Migration