Help Center/ GaussDB/ Feature Guide(Centralized_8.x)/ Partitioned Table/ Partitioned Table O&M Management
Updated on 2024-06-07 GMT+08:00
View PDF
Share

Partitioned Table O&M Management

Partitioned table O&M management includes partition management, partitioned table management, partitioned index management, and partitioned table statement concurrency support.

  • Partition management: also known as partition-level DDL operations, including ADD, DROP, EXCHANGE, TRUNCATE, SPLIT, MERGE, MOVE, and RENAME.
  • For hash partitions, operations involving partition quantity change will cause data re-shuffling, including ADD, DROP, SPLIT, and MERGE. Therefore, GaussDB does not support these operations.
  • Operations involving partition data change will invalidate global indexes, including DROP, EXCHANGE, TRUNCATE, SPLIT, and MERGE. You can use the UPDATE GLOBAL INDEX clause to update global indexes synchronously.
  • Most partition DDL operations use PARTITION/SUBPARTITION and PARTITION/SUBPARTITION FOR to specify partitions. For PARTITION/SUBPARTITION, you need to specify the partition name. For PARTITION/SUBPARTITION FOR, you need to specify any partition value within the partition range. For example, if the range of partition part1 is defined as [100, 200), partition part1 and partition for(150) function the same.
  • The DDL execution cost varies depending on the partition. The target partition will be locked during DDL execution. Therefore, you need to evaluate the cost and impact on services. Generally, the execution cost of splitting and merging is much greater than that of other partition DDL operations and is positively correlated with the size of the source partition. The cost of exchanging is mainly caused by global index rebuilding and validation. The cost of moving is limited by disk I/O. The execution cost of other partition DDL operations is low.
  • Partitioned table management: In addition to the functions inherited from ordinary tables, you can enable or disable row migration for partitioned tables.
  • Partitioned index management: You can invalidate indexes or index partitions or rebuild invalid indexes or index partitions. For example, global indexes become invalid due to partition management operations.
  • Partitioned table statement concurrency support: When partition-level DDL operations and partition-level DQL/DML operations are applied to different partitions, concurrency at the execution layer is supported.