StarRocks

Bruin supports StarRocks as a native SQL data platform through its MySQL-compatible FE query port. This lets you build and materialize StarRocks tables, run data quality checks, and use sensors — in addition to using StarRocks as an ingestion source or destination.

StarRocks needs specific DDL to create tables (a table model, a DISTRIBUTED BY clause, PROPERTIES, and optionally PARTITION BY), which the generic MySQL adapter does not emit — the starrocks.* asset types below generate it.

Connection

Add a StarRocks entry under connections in .bruin.yml. The same connection is used for both native assets and ingestion.

connections:
  starrocks:
    - name: "starrocks-default"
      username: "root"
      host: "starrocks.example.com"
      port: 9030                    # optional, defaults to 9030 (FE MySQL protocol port)
      password: "XXXXXXXXXX"        # optional
      database: "analytics"         # optional
      ssl: "true"                   # optional — "true" or "skip-verify"

StarRocks Assets

starrocks.sql

Executes a materialized StarRocks SQL asset.

/* @bruin
name: analytics.example
type: starrocks.sql
materialization:
    type: table
@bruin */

SELECT
    id,
    country,
    name
FROM staging.customers

Supported table materialization strategies:

• create+replace
• append
• delete+insert
• merge
• truncate+insert
• time_interval
• ddl

View materialization is also supported. For local and single-node StarRocks clusters, Bruin creates StarRocks-managed tables with PROPERTIES ("replication_num" = "1"). Atomic replacements (create+replace, delete+insert, truncate+insert, seeds) use StarRocks' ALTER TABLE ... SWAP WITH ....

Table layout

Distribution and partitioning are taken from the standard materialization fields, so they work the same way as on the other platforms:

• materialization.cluster_by → DISTRIBUTED BY HASH(...) (defaults to the key columns)
• materialization.partition_by → PARTITION BY (...) (a column or expression such as date_trunc('day', event_date))

partition_by is emitted verbatim so that expressions work, so it is not automatically backtick-quoted. If you partition on a single column whose name is a StarRocks reserved word (e.g. date, value, key), quote it yourself: partition_by: "date".

StarRocks-specific layout that has no materialization equivalent is declared under starrocks:

materialization:
  type: table
  strategy: create+replace
  cluster_by: [account_id]          # DISTRIBUTED BY HASH(account_id)
  partition_by: event_date          # PARTITION BY (event_date)

starrocks:
  table_model: primary_key          # duplicate_key | unique_key | primary_key
  buckets: 8                         # defaults to 1
  properties:
    replication_num: "1"

When any of these are set (or columns are declared), Bruin emits a typed CREATE TABLE with the key clause, PARTITION BY, DISTRIBUTED BY HASH(...) BUCKETS, and PROPERTIES. Otherwise it falls back to CREATE TABLE ... AS SELECT.

Merge materialization

StarRocks has no MERGE INTO statement. Bruin implements merge with a StarRocks PRIMARY KEY table: matching rows are replaced and new rows inserted by a plain INSERT. Bruin infers starrocks.table_model: primary_key for merge assets, creates the table if it does not exist, and upserts on the primary key.

Merge assets must declare typed columns and at least one primary_key column. Per-column merge_sql expressions are not supported (StarRocks upserts whole rows on the primary key) — encode that logic in the asset query instead.

/* @bruin
name: analytics.accounts
type: starrocks.sql
materialization:
    type: table
    strategy: merge

columns:
  - name: account_id
    type: BIGINT
    primary_key: true
  - name: status
    type: VARCHAR(32)
    update_on_merge: true
@bruin */

SELECT account_id, status
FROM staging.accounts

starrocks.seed

Loads a local CSV file into StarRocks.

name: analytics.seed_contacts
type: starrocks.seed

columns:
  - name: name
    type: STRING
  - name: channel
    type: STRING

parameters:
  path: seed.csv
  file_type: csv

starrocks.sensor.table

Waits until a StarRocks table exists.

name: analytics.wait_for_daily_summary
type: starrocks.sensor.table
parameters:
    table: analytics.daily_summary
    poke_interval: 30

starrocks.sensor.query

Waits until a StarRocks query returns at least one row.

name: analytics.wait_for_orders
type: starrocks.sensor.query
parameters:
    query: SELECT 1 FROM analytics.orders WHERE order_date = "{{ end_date }}" LIMIT 1

starrocks.source

Defines an existing StarRocks table as a source asset for lineage and documentation.

name: analytics.raw_orders
type: starrocks.source

columns:
  - name: order_id
    type: BIGINT