Skip to main content
Version: Next

Manage Table Partitions

Introduction​

Although many catalogs inherently manage partitions automatically, there are scenarios where manual partition management is necessary. Usage scenarios like managing the TTL (Time-To-Live) of partition data, gathering statistics on partition metadata, and optimizing queries through partition pruning. For these reasons, Apache Gravitino provides capabilities of partition management.

Requirements and Limitations​

  • Partition management is based on the partitioned table, so ensure you are operating on a partitioned table.

The following table shows the partition operations supported across various catalogs in Gravitino:

OperationHive catalogIceberg catalogJdbc-MySQL catalogJdbc-PostgreSQL catalogJdbc-Doris catalog
Add Partition✔✘✘✘✔
Get Partition by Name✔✘✘✘✔
List Partition Names✔✘✘✘✔
List Partitions✔✘✘✘✔
Drop Partition✔✘✘✘✔
WELCOME FEEDBACK

If you need additional partition management support for a specific catalog, create an issue on the Gravitino repository.

Partition Operations​

Add Partition​

Match the partition types you want to add with the table's partitioning types; Gravitino supports adding the following partition types:

Partition TypeDescription
identityAn identity partition represents a result of identity partitioning.
rangeA range partition represents a result of range partitioning.
listA list partition represents a result of list partitioning.

For JSON examples:

{
  "type": "identity",
  "name": "dt=2008-08-08/country=us",
  "fieldNames": [
    [
      "dt"
    ],
    [
      "country"
    ]
  ],
  "values": [
    {
      "type": "literal",
      "dataType": "date",
      "value": "2008-08-08"
    },
    {
      "type": "literal",
      "dataType": "string",
      "value": "us"
    }
  ]
}
note

The values of the field values must be the same ordering as the values of fieldNames.

When adding an identity partition to a partitioned Hive table, the specified partition name is ignored. This is because Hive generates the partition name based on field names and values.

For Java examples:

Partition partition =
        Partitions.identity(
            "dt=2008-08-08/country=us",
            new String[][] {{"dt"}, {"country"}},
            new Literal[] {
              Literals.dateLiteral(LocalDate.parse("2008-08-08")), Literals.stringLiteral("us")
            },
            Maps.newHashMap());
note

The values are in the same order as the field names.

When adding an identity partition to a partitioned Hive table, the specified partition name is ignored. This is because Hive generates the partition name based on field names and values.

Add a partition to a partitioned table by sending a POST request to the /api/metalakes/{metalake_name}/catalogs/{catalog_name}/schemas/{schema_name}/tables/{partitioned_table_name}/partitions endpoint or by using the Gravitino Java client. The following is an example of adding an identity partition to a Hive partitioned table:

curl -X POST -H "Accept: application/vnd.gravitino.v1+json" \
-H "Content-Type: application/json" -d '{
  "partitions": [
    {
      "type": "identity",
      "fieldNames": [
        [
          "dt"
        ],
        [
          "country"
        ]
      ],
      "values": [
        {
          "type": "literal",
          "dataType": "date",
          "value": "2008-08-08"
        },
        {
          "type": "literal",
          "dataType": "string",
          "value": "us"
        }
      ]
    }
  ]
}' http://localhost:8090/api/metalakes/metalake/catalogs/catalog/schemas/schema/tables/table/partitions

Get a Partition by Name​

Get a partition by its name via sending a GET request to the /api/metalakes/{metalake_name}/catalogs/{catalog_name}/schemas/{schema_name}/tables/{partitioned_table_name}/partitions/{partition_name} endpoint or by using the Gravitino Java client. The following is an example of getting a partition by its name:

curl -X GET -H "Accept: application/vnd.gravitino.v1+json" \
-H "Content-Type: application/json" \
http://localhost:8090/api/metalakes/metalake/catalogs/catalog/schemas/schema/tables/table/partitions/p20200321
tip

If the partition name contains special characters, you should use URL encoding. For example, if the partition name is dt=2008-08-08/country=us you should use dt%3D2008-08-08%2Fcountry%3Dus in the URL.

List Partition Names Under a Partitioned Table​

List all partition names under a partitioned table by sending a GET request to the /api/metalakes/{metalake_name}/catalogs/{catalog_name}/schemas/{schema_name}/tables/{partitioned_table_name}/partitions endpoint or by using the Gravitino Java client. The following is an example of listing all the partition names under a partitioned table:

curl -X GET -H "Accept: application/vnd.gravitino.v1+json" \
-H "Content-Type: application/json" \
http://localhost:8090/api/metalakes/metalake/catalogs/catalog/schemas/schema/tables/table/partitions

List Partitions Under a Partitioned Table​

If you want to get more detailed information about the partitions under a partitioned table, you can list all partitions under a partitioned table by sending a GET request to the /api/metalakes/{metalake_name}/catalogs/{catalog_name}/schemas/{schema_name}/tables/{partitioned_table_name}/partitions endpoint or by using the Gravitino Java client. The following is an example of listing all the partitions under a partitioned table:

curl -X GET -H "Accept: application/vnd.gravitino.v1+json" \
-H "Content-Type: application/json" \
http://localhost:8090/api/metalakes/metalake/catalogs/catalog/schemas/schema/tables/table/partitions?details=true

Drop a Partition by Name​

Drop a partition by its name via sending a DELETE request to the /api/metalakes/{metalake_name}/catalogs/{catalog_name}/schemas/{schema_name}/tables/{partitioned_table_name}/partitions/{partition_name} endpoint or by using the Gravitino Java client. The following is an example of dropping a partition by its name:

curl -X DELETE -H "Accept: application/vnd.gravitino.v1+json" \
-H "Content-Type: application/json" \
http://localhost:8090/api/metalakes/metalake/catalogs/catalog/schemas/schema/tables/table/partitions/p20200321
tip

If the partition name contains special characters, you should use URL encoding. For example, if the partition name is dt=2008-08-08/country=us you should use dt%3D2008-08-08%2Fcountry%3Dus in the URL.