Hologres:Time Travel

Set Time Travel

• Syntax:

  To enable Time Travel, set the following two properties when you create a table:

  CREATE TABLE <table_name> (...)
  WITH (
      enable_mvcc = 'true|false',
      mvcc_gc_ttl_seconds = '<num>'
  );

• Notes

  • Set these parameters when creating a table. They cannot be configured on an existing one.

  • Enabling Time Travel stores multiple data versions, which increases storage usage.

  • Enabling Time Travel can slightly degrade query performance.

• Parameters

  Parameter	Description
  enable_mvcc	Specifies whether to enable Time Travel. Valid values: true: Enables Time Travel. This setting cannot be disabled once enabled. false (default): Disables Time Travel.
  mvcc_gc_ttl_seconds	The Time to Live (TTL) for each data version, in seconds (s). The default value is 86400, which is one day (24 * 3600 s). The minimum value is 0. This does not disable Time Travel. Instead, it means that data versions are recycled after 0 seconds and cannot be queried. Note Hologres periodically deletes expired data versions. Consequently, queries for data beyond the specified TTL will fail. An example error is: Tablexxx only supports time traveling between xxx and xxx;.

Query data using Time Travel

After you enable Time Travel, query data from a specific point in time within the TTL period using the FOR TIMESTAMP AS OF clause. The syntax is as follows:

SELECT * FROM {table_name} FOR TIMESTAMP AS OF (now() - interval '1 hour');
SELECT * FROM {table_name} FOR TIMESTAMP AS OF '2025-08-05 11:18:50+08';
SELECT * FROM {table_name} FOR TIMESTAMP AS OF 1754374966;

Modify Time Travel TTL

Modify the TTL for Time Travel. The syntax is as follows:

ALTER TABLE {table_name} SET (mvcc_gc_ttl_seconds =<num>);

Modifying the TTL has the following effects:

• When you decrease the TTL: Hologres deletes expired data versions in the background under the new TTL. Queries for data beyond the new TTL will fail.

• When you increase the TTL: Data versions deleted under the previous, shorter TTL cannot be recovered. To avoid potential data gaps and inaccurate queries, do not increase the TTL in a production environment.