Snowflake source

tap-snowflake extracts Snowflake tables through the standard Singer path.

Support

Source	Status	Replication
Snowflake	Experimental	FULL_TABLE and INCREMENTAL; no FastSync

Prerequisites

The Snowflake role needs warehouse usage, database and schema usage, and SELECT on every replicated table. Choose a warehouse with enough capacity for the configured query cadence.

Configuration

id: "snowflake_source"
name: "Operational Snowflake"
type: "tap-snowflake"
owner: "data-platform@example.com"
db_conn:
  account: "<ACCOUNT>"
  dbname: "<DATABASE>"
  user: "<USER>"
  password: "{{ env_var['SNOWFLAKE_PASSWORD'] }}"
  warehouse: "<WAREHOUSE>"
target: "postgres_dwh"
batch_size_rows: 20000
stream_buffer_size: 0
schemas:
  - source_schema: "PUBLIC"
    target_schema: "repl_snowflake"
    tables:
      - table_name: "ORDERS"
        replication_method: "INCREMENTAL"
        replication_key: "UPDATED_AT"

INCREMENTAL requires a stable key whose new value is greater than or equal to the saved bookmark. It does not capture deletes. Use FULL_TABLE when no suitable key exists and account for the cost of scanning the entire source table on every run.

Common settings are documented in YAML configuration; generate the full template with pipelinewise init.