All Products
Search

This Product

    MaxCompute:CDC (invitational preview)

    Document Center

    MaxCompute:CDC (invitational preview)

    Last Updated:Aug 14, 2026

    Change Data Capture (CDC) records row-level inserts, updates, and deletes on Delta tables as an incremental event stream. Downstream systems can consume this stream for incremental computing, data synchronization, stream processing, and auditing — without scanning the full table on each run.

    CDC is currently in invitational preview. For usage instructions, see Incremental calculation overview.

    Use cases

    • Incremental computing: Consume change records to update materialized views without reprocessing the full table.

    • Stream processing: Feed CDC records into Flink jobs or other stream processors.

    • Multi-engine data synchronization: Propagate incremental changes across different compute engines.

    • Log auditing: Retain a complete record of all row-level operations for compliance and governance.

    Choose a CDC mode

    CDC supports two modes. Choose based on whether your workload requires real-time writes via Tunnel.

    Synchronous CDC

    Asynchronous CDC

    SQL DML operations

    Supported

    Supported

    Real-time writes via Tunnel

    Not supported

    Supported

    CDC data availability

    Immediately after the SQL operation completes

    Asynchronously, after the configured interval

    Enable CDC on a Delta table

    Synchronous CDC

    Add "acid.cdc.mode.enable"="true" and "acid.cdc.build.async" = "false" properties when creating a Delta table.

    CREATE TABLE acid_with_cdc_tbl (pk BIGINT NOT NULL PRIMARY KEY, val BIGINT) 

tblproperties ("transactional" = "true", 
               "acid.cdc.mode.enable"="true",
               "acid.cdc.build.async" = "false");
    Note

    • acid.cdc.mode.enable: After the Delta Table CDC feature is enabled, it runs asynchronously by default. CDC is built asynchronously when data changes are triggered by SQL

      DML operations, and CDC generation is also supported asynchronously in Tunnel real-time write scenarios.

    • acid.cdc.build.async: Controls how CDC data is built. When set to "false", it creates a synchronous CDC table, and CDC data is generated immediately after SQL DML operations complete. Real-time writes through Tunnel are not supported.

    • cdc.data.retain.hours: An optional property. Retention period for CDC data, in hours. Range: 1–168. Default value: 24.

    • Tables with CDC enabled do not support the INSERT INTO operation. We recommend that you use MERGE INTO instead.

    Optional properties:

    Property

    Description

    Default

    "cdc.data.retain.hours"="24"

    Retention period for CDC data, in hours. Range: 1–168.

    24

    Asynchronous CDC

    Add "acid.cdc.mode.enable"="true", "acid.cdc.build.async"="true", and "acid.cdc.build.interval"="300" when creating a Delta table. You can optionally add the "cdc.data.retain.hours"="24" property.

    CREATE TABLE acid_with_cdc_build_tbl (pk BIGINT NOT NULL PRIMARY KEY, val BIGINT)
tblproperties ("transactional" = "true",
               "acid.cdc.mode.enable"="true",
               "acid.cdc.build.async"="true",
               "acid.cdc.build.interval"="300");
    Note

    • acid.cdc.mode.enable: After the Delta Table CDC feature is enabled, it runs asynchronously by default. CDC is built asynchronously when data changes are triggered by SQL

      DML operations, and CDC generation is also supported asynchronously in Tunnel real-time write scenarios.

    • acid.cdc.build.async: Enables asynchronous CDC building. Real-time writes to the table through Tunnel are supported, and CDC data is also generated asynchronously for SQL DML operations.

    • acid.cdc.build.interval: The interval for asynchronous building, in seconds. Valid values: [60-3540]. Configure this parameter based on business or incremental scenarios.

    • cdc.data.retain.hours: An optional property. Retention period for CDC data, in hours. Range: 1–168. Default value: 24.

    • Other optional parameters (project level or session level): odps.storage.orc.enable.memcmp.sort.key="true", We recommend enabling it at the project level, which improves the performance of both asynchronous CDC building and queries.

    Query CDC data

    Use the table_changes function to query CDC change records by version or timestamp range.

    Syntax

    SELECT * FROM table_changes('<table_name>', <start> [, <end>]);

    Parameters

    Parameter

    Required

    Type

    Description

    table_name

    Yes

    —

    The Delta table to query.

    start

    Yes

    BIGINT or STRING

    Starting version for the CDC data query. Find version numbers with SHOW HISTORY FOR TABLE <table_name>. For STRING, use yyyy-mm-dd hh:mi:ss.

    end

    No

    BIGINT or STRING

    Ending version for the CDC data query. Defaults to the latest version if omitted. For STRING, use yyyy-mm-dd hh:mi:ss.

    Return columns

    In addition to the table's data columns, table_changes returns three system columns:

    Column

    Description

    __meta_timestamp

    System time when the row was written.

    __meta_op_type

    Operation type: 1 = INSERT, 0 = DELETE.

    __meta_is_update

    Whether the row is part of an UPDATE: 1 = yes, 0 = no.

    Combine __meta_op_type and __meta_is_update to identify the full operation:

    __meta_op_type

    __meta_is_update

    Operation

    1

    0

    New row from an INSERT

    1

    1

    Post-update value (the new value after UPDATE)

    0

    1

    Pre-update value (the original value before UPDATE)

    0

    0

    Deleted row

    Example

    Step 1: Create a CDC-enabled table.

    CREATE TABLE acid_cdc_table(id1 STRING NOT NULL, id2 STRING NOT NULL, key1 BIGINT, key2 BIGINT, PRIMARY KEY(id1, id2))
tblproperties("transactional" = "true", "acid.cdc.mode.enable"="true");

    Step 2: Insert data.

    -- Data insertion time 2025-04-07 11:56:57
INSERT INTO acid_cdc_table VALUES ('1', '1006', 1006, 1006);
-- Data insertion time 2025-04-07 12:15:00
INSERT INTO acid_cdc_table VALUES ('1', '1008', 1008, 1008);
-- Data insertion time 2025-04-07 13:24:00
INSERT INTO acid_cdc_table VALUES ('1', '1032', 1032, 1032);
-- Data insertion time 2025-04-07 14:00:00
INSERT INTO acid_cdc_table VALUES ('1', '1045', 1045, 1045);
-- Data insertion time 2025-04-07 14:47:00
INSERT INTO acid_cdc_table VALUES ('1', '1045', 1045, 1045);

    Step 3: Find table versions.

    SHOW HISTORY FOR TABLE acid_cdc_table;

    Output:

    ObjectType      ObjectId                                ObjectName              VERSION(LSN)            Time                    Operation
TABLE           a4a78d3f6af04d85a57a90deff884021        acid_cdc_table          0000000000000001        2025-04-07 11:55:59     CREATE
TABLE           a4a78d3f6af04d85a57a90deff884021        acid_cdc_table          0000000000000002        2025-04-07 11:56:57     APPEND
TABLE           a4a78d3f6af04d85a57a90deff884021        acid_cdc_table          0000000000000003        2025-04-07 12:00:13     MINOR_COMPACT
TABLE           a4a78d3f6af04d85a57a90deff884021        acid_cdc_table          0000000000000004        2025-04-07 12:15:32     APPEND
TABLE           a4a78d3f6af04d85a57a90deff884021        acid_cdc_table          0000000000000005        2025-04-07 12:30:02     MINOR_COMPACT
TABLE           a4a78d3f6af04d85a57a90deff884021        acid_cdc_table          0000000000000006        2025-04-07 13:24:47     APPEND
TABLE           a4a78d3f6af04d85a57a90deff884021        acid_cdc_table          0000000000000007        2025-04-07 13:30:02     MINOR_COMPACT
TABLE           a4a78d3f6af04d85a57a90deff884021        acid_cdc_table          0000000000000008        2025-04-07 14:00:41     APPEND
TABLE           a4a78d3f6af04d85a57a90deff884021        acid_cdc_table          0000000000000009        2025-04-07 14:15:15     MINOR_COMPACT
TABLE           a4a78d3f6af04d85a57a90deff884021        acid_cdc_table          0000000000000010        2025-04-07 14:47:46     APPEND
TABLE           a4a78d3f6af04d85a57a90deff884021        acid_cdc_table          0000000000000011        2025-04-07 15:00:11     MINOR_COMPACT

    Step 4: Query CDC records.

    Query all changes after 2025-04-07 12:00:00 (equivalent to version 3):

    SELECT * FROM table_changes('acid_cdc_table', '2025-04-07 12:00:00');
-- Equivalent to
SELECT * FROM table_changes('acid_cdc_table', 3);

    Output:

    +------------+------------+------------+------------+------------------+----------------+------------------+
| id1        | id2        | key1       | key2       | __meta_timestamp | __meta_op_type | __meta_is_update |
+------------+------------+------------+------------+------------------+----------------+------------------+
| 1          | 1045       | 1045       | 1045       | 2025-04-07 14:00:34 | 1           | 0                |
| 1          | 1008       | 1008       | 1008       | 2025-04-07 12:15:28 | 1           | 0                |
| 1          | 1032       | 1032       | 1032       | 2025-04-07 13:24:43 | 1           | 0                |
| 2          | 1045       | 1045       | 1045       | 2025-04-07 14:47:41 | 1           | 0                |
+------------+------------+------------+------------+------------------+----------------+------------------+

    Query changes within a specific range (2025-04-07 12:00:00 to 13:30:00, equivalent to versions 3–6):

    SELECT * FROM table_changes('acid_cdc_table', '2025-04-07 12:00:00', '2025-04-07 13:30:00');
-- Equivalent to
SELECT * FROM table_changes('acid_cdc_table', 3, 6);

    Output:

    +------------+------------+------------+------------+------------------+----------------+------------------+
| id1        | id2        | key1       | key2       | __meta_timestamp | __meta_op_type | __meta_is_update |
+------------+------------+------------+------------+------------------+----------------+------------------+
| 1          | 1008       | 1008       | 1008       | 2025-04-07 12:15:28 | 1           | 0                |
| 1          | 1032       | 1032       | 1032       | 2025-04-07 13:24:43 | 1           | 0                |
+------------+------------+------------+------------+------------------+----------------+------------------+

    Consume CDC data with a Stream

    A Stream object tracks a read offset into a Delta table's CDC history. Create a Stream with "read_mode"="cdc" to consume CDC data incrementally.

    For full Stream documentation, see Stream object.

    Syntax

    CREATE STREAM [IF NOT EXISTS] <stream_name>
ON TABLE <delta_table_name> VERSION AS OF <v>
strmproperties ("read_mode"="cdc")

    The "read_mode"="cdc" property sets the Stream to consume CDC data based on the query version range.

    Example

    -- Create the source table with CDC enabled.
CREATE TABLE acid_with_cdc_stream (id1 BIGINT NOT NULL PRIMARY KEY, id2 BIGINT)
tblproperties ("transactional" = "true", "acid.cdc.mode.enable"="true");

-- Insert data.
INSERT INTO acid_with_cdc_stream VALUES (1, 1006), (2, 1008), (3, 1032);

-- Create a Stream starting at version 1.
CREATE STREAM delta_table_stream ON TABLE acid_with_cdc_stream VERSION AS OF 1 strmproperties ("read_mode"="cdc");

-- Inspect the Stream.
DESC STREAM delta_table_stream;

    Output:

    Name                                    delta_table_stream
Project                                 yunqi_y****
Schema                                  default
Create Time                             2024-12-03 11:13:12
Last Modified Time                      2024-12-03 11:13:12
Offset Version                          1
Reference Table Project                 yunqi_y****
Reference Table Schema                  default
Reference Table Name                    acid_with_cdc_stream
Reference Table Id                      b89ec113f50944d5b8e52ce6a00c****
Reference Table Version                 2
Parameters                              {"read_mode": "cdc"}

    What's next