Hi team, is it not possible to load rows with different columns using the Starrocks Kafka connector? We have a Primary Key table and we would like to stream load updates to different columns. However, if we set

partial_update: true

, we see an error like

com.starrocks.data.load.stream.exception.StreamLoadFailException: Stream load failed because of error, db: dev, table: trait_data, label: -fd20516a-77cf-4f5d-906f-4c11a13bfc90,
responseBody: {
    "Status": "ANALYSIS_ERROR",
    "Message": "key column workspace_id not in partial update columns"
}

the message we are sending does have the primary key here - wondering if we need to specify the column list as some sink property? But there's nothing in the docs about this, and doing this would prevent us from updating different columns

You can use partial update using starrocks Kafka connector. Can you share

show create table dev.trait_data;

Are you using it on distributed mode and how did you specify which columns to update?

Hmm actually we're seeing this issue with both routine load and kafka connector 😕 Table:

CREATE TABLE `trait_data` (
  `tenant_id` varchar(36) NOT NULL COMMENT "",
  `object_id` varchar(36) NOT NULL COMMENT "",
  `entity_id` varchar(32) NOT NULL COMMENT "",
  `refreshed_at` datetime NOT NULL COMMENT "",
  `trait_0` varchar(1048576) NULL COMMENT "",
  `trait_1` varchar(1048576) NULL COMMENT "",
  `trait_2` varchar(1048576) NULL COMMENT "",
  `trait_3` varchar(1048576) NULL COMMENT "",
  `trait_4` varchar(1048576) NULL COMMENT "",
  `trait_5` varchar(1048576) NULL COMMENT "",
  `trait_6` varchar(1048576) NULL COMMENT "",
  `trait_7` varchar(1048576) NULL COMMENT "",
  `trait_8` varchar(1048576) NULL COMMENT "",
  `trait_9` varchar(1048576) NULL COMMENT "",
) ENGINE=OLAP 
PRIMARY KEY(`tenant_id`, `object_id`, `entity_id`)
PARTITION BY (`tenant_id`,`object_id`)
DISTRIBUTED BY HASH(`tenant_id`, `object_id`, `entity_id`)
ORDER BY(`tenant_id`, `object_id`, `entity_id`)
PROPERTIES (
"compression" = "LZ4",
"enable_persistent_index" = "true",
"fast_schema_evolution" = "true",
"replicated_storage" = "true",
"replication_num" = "1"
);

Routine load:

CREATE ROUTINE LOAD dev.trait_updates on trait_data
COLUMNS (
	tenant_id, 
	object_id, 
	entity_id, 
	refreshed_at, 
	trait_0, 
	trait_1, 
	trait_2, 
	trait_3, 
	trait_4, 
	trait_5, 
	trait_6, 
	trait_7, 
	trait_8, 
	trait_9
)
PROPERTIES
(
"format"="json",
"partial_update"="true",
)
FROM KAFKA
(
"kafka_broker_list"="<...>:9092",
"kafka_topic"="trait_updates",
);

Example kafka message:

{
  "object_id": "b54fcf32-472a-413c-bb5c-a7ae2705649f",
  "entity_id": "b54f_DCE40A21252F1457BBD84",
  "refreshed_at": "2025-11-07T20:00:31.625536406Z",
  "trait_1": "1",
  "tenant_id": "77449ebe-061d-4071-837e-e07b54f00ee2"
}

This loads successfully, but erases all other columns