PolarDB:DDL replication

Supported DDL types

The following DDL statements are not replicated and must be applied separately to each cluster: CCI-related statements (CREATE COLUMNAR INDEX, DROP COLUMNAR INDEX, ALTER COLUMNAR INDEX), SET GLOBAL, and SQL throttling rules. For details, see the Remarks column.

Table groups

PolarDB-X supports two types of table groups: explicit and implicit.

Explicit table groups are managed directly with DDL statements:

• Create: CREATE TABLEGROUP

• Delete: DROP TABLEGROUP

• Modify: ALTER TABLEGROUP

For more information, see Table groups.

Implicit table groups are created automatically by PolarDB-X when you create, drop, or repartition a table. They are named sequentially (tg1, tg2, tg3, ...), but the sequence numbers are non-deterministic — parallel table creation produces random assignments, and rolled-back DDL tasks cause sequence numbers to be skipped.

This non-determinism creates a replication challenge: if the primary and secondary clusters independently generate implicit table groups, the same table may be assigned different table group names across clusters. To solve this, PolarDB-X extends DDL statements with explicit table group bindings before writing them to the binary log. Secondary clusters replay the extended statements and use the same table group names as the primary cluster.

Example: `CREATE TABLE`

Original statement submitted on the primary cluster:

CREATE TABLE IF NOT EXISTS tb1 (
  a int PRIMARY KEY,
  b int,
  c int,
  d varchar(10) UNIQUE,
  INDEX b(b),
  INDEX b_2(b),
  KEY b_3 (b),
  KEY b_4 (b),
  UNIQUE KEY b_5 (b),
  UNIQUE KEY b_6 (b),
  UNIQUE INDEX b_7 (b),
  UNIQUE INDEX b_8 (b),
  INDEX g1(b),
  KEY g2 (b),
  UNIQUE KEY g3 (b),
  UNIQUE INDEX g4 (b)
) DEFAULT CHARACTER SET = utf8mb4 DEFAULT COLLATE = utf8mb4_general_ci;

Example: `ALTER TABLE ... MERGE PARTITIONS`

Original statement submitted on the primary cluster:

ALTER TABLE tb1 MERGE PARTITIONS p1, p2 TO p12;

Extended statement recorded in the binary log:

ALTER TABLE tb1 MERGE PARTITIONS p1, p2 TO p12 WITH TABLEGROUP=tg29 IMPLICIT;

The WITH TABLEGROUP=tg29 IMPLICIT clause pins the implicit table group name so secondary clusters use the same name instead of generating their own.

Multi-stream replication

In single-stream replication, DDL statements are replicated one at a time, strictly following binary log order.

Multi-stream replication runs multiple replication links in parallel to increase throughput. This introduces a coordination problem for DDL: if one stream applies a DDL statement immediately upon receiving it, the schema becomes temporarily inconsistent with the other streams that haven't received it yet. Any DML arriving on those streams would run against the wrong schema, causing data errors or exceptions.

To prevent this, PolarDB-X includes a distributed DDL replication engine. After a DDL statement arrives on any stream, all streams pause and wait until every stream has received the statement. Only then does each stream apply the DDL and resume DML processing. This coordination is handled by an efficient multi-link coordination algorithm built into the engine.

Limitations

CREATE TABLE test_pg3 (
  id int
) DEFAULT CHARSET = utf8mb4 DEFAULT COLLATE = utf8mb4_general_ci
PARTITION BY RANGE (id) (
  PARTITION p0 VALUES LESS THAN (1000) LOCALITY 'dn=xdevelop-240524092100-31ef-bngl-dn-1',
  PARTITION p1 VALUES LESS THAN (2000) LOCALITY 'dn=xdevelop-240524092100-31ef-bngl-dn-1'
);

CREATE TABLE test_pg3 (
  id int
) DEFAULT CHARACTER SET = utf8mb4 DEFAULT COLLATE = utf8mb4_general_ci
PARTITION BY RANGE (id) (
  PARTITION p0 VALUES LESS THAN (1000),
  PARTITION p1 VALUES LESS THAN (2000)
) WITH TABLEGROUP = tg3778 IMPLICIT;

ALTER TABLE special_dml_test1
  MOVE PARTITIONS (p2, p4, p6, p8) TO 'xdevelop-240524092100-31ef-bngl-dn-0';

ALTER TABLE special_dml_test1;