Serverless StarRocks access to Fluss (Beta)

Step 1: Create a Fluss catalog

1. Log on to the EMR Serverless StarRocks console. On the instance list page, click Connect Instance in the Actions column of the target instance. For details, see Connect to a StarRocks instance through EMR StarRocks Manager.

2. In the SQL Editor, execute the following SQL statement to register Fluss as an external catalog:

   CREATE EXTERNAL CATALOG <catalog_name>
   PROPERTIES (
       'type' = 'fluss',
       'bootstrap.servers' = '<Fluss cluster bootstrap address>',
       'fluss.option.client.security.protocol' = 'SASL',
       'fluss.option.client.security.sasl.mechanism' = 'PLAIN',
       'fluss.option.client.security.sasl.username' = '<Fluss username>',
       'fluss.option.client.security.sasl.password' = '<Fluss password>'
   );

Parameter description

After the catalog is created, you can query it directly using the three-part name (Catalog.Database.Table) without switching context. The default Fluss database name is fluss:

SELECT * FROM <catalog_name>.<database_name>.<table_name>;

You can also switch to the Fluss catalog before querying:

-- Switch catalog and database
SET CATALOG <catalog_name>;
USE <db_name>;

-- Query the target table
SELECT COUNT(*) FROM <table_name> LIMIT 10;

Or, do it in one step:

USE <catalog_name>.<db_name>;

Step 2: Specify the query mode using a suffix

For Fluss tables that have enabled the lakehouse-stream integration (table.datalake.enabled = 'true' is set at table creation), data exists in both the Fluss cluster (real-time hot data) and the Paimon data lake (tiered historical data). StarRocks supports specifying the query mode by adding a suffix to the table name, enabling flexible access to data at different layers. The query modes are:

Consider an online sales order table order, which records user order data in real time (including fields such as order ID, order amount, order time, and region) and has enabled the lakehouse-stream integration. The following examples show typical business scenarios for the three query modes:

-- Scenario 1: Real-time monitoring of today's order count and amount ($rt, second-level latency)
SELECT COUNT(*) AS order_cnt, SUM(order_amount) AS total_amount
FROM `fluss_catalog`.`fluss`.order$rt
WHERE order_time >= CURRENT_DATE;

-- Scenario 2: Calculate week-over-week sales growth rate ($lake, query Paimon historical data)
SELECT
    this_week AS this_week_sales,
    last_week AS last_week_sales,
    ROUND((this_week - last_week) * 100.0 / NULLIF(last_week, 0), 2) AS wow_growth_rate
FROM (
    SELECT
        SUM(CASE WHEN YEARWEEK(order_time) = YEARWEEK(CURRENT_DATE)     THEN order_amount ELSE 0 END) AS this_week,
        SUM(CASE WHEN YEARWEEK(order_time) = YEARWEEK(CURRENT_DATE) - 1 THEN order_amount ELSE 0 END) AS last_week
    FROM `fluss_catalog`.`fluss`.order$lake
) t;

-- Scenario 3: Query total order amount by region in the past 7 days (no suffix, Union Read merges hot and cold data)
SELECT region, SUM(order_amount) AS total_amount
FROM `fluss_catalog`.`fluss`.order
WHERE order_time >= CURRENT_DATE - INTERVAL '7' DAY
GROUP BY region
ORDER BY total_amount DESC;

View created catalogs

-- View all catalogs
SHOW CATALOGS;

-- View the CREATE statement of a specified catalog
SHOW CREATE CATALOG <catalog_name>;

View the Fluss table schema

DESCRIBE <catalog_name>.<database_name>.<table_name>;

View the Fluss database list

SHOW DATABASES FROM <catalog_name>;

Drop the Fluss catalog

Dropping a catalog only removes the mapping in StarRocks. The actual data in Fluss is not affected.

DROP CATALOG <catalog_name>;