writing directly to workers create issues with materialized tables

[dev-mraj]

Hi,

We use this query to ingest large amount of data in batches to write workers directly (not servers) using http api (8123)

INSERT INTO db_abc.log Native SETTINGS prefer_cnch_catalog=1, max_threads_for_cnch_dump = 8

but get this errors that db_abc.log_attr_index table do not exists

CREATE MATERIALIZED VIEW IF NOT EXISTS 
				db_abc.log_attr_r_idx_mv 
				TO db_abc.log_attr_index 
			AS
			SELECT
				arrayJoin(mapKeys(resource_attributes)) AS key,
				resource_attributes[key] AS value,
				'resource' AS kind,
				toUnixTimestamp64Milli(now64()) AS ts
			FROM
			(
				SELECT resource_attributes
				FROM db_abc.log
				GROUP BY type,
					resource_attributes,
					timestampMs
			)
			GROUP BY key, value

[kevinthfang]

Did you write to the workers using jdbc?

[dev-mraj]

using using http api (port 8123) so client is not using jdbc its based on golang clickhouse driver and ch-go

its binary blocks works fine when sending to servers but not with workers, and reason behind sending to worker is scaling capability because servers can't scale horizontally (afaik) for any given single table

[jenrryyou]

can you show the DDL for table db_abc.log