Google Bigtable as a wide-column NoSQL source or target. One connection maps to one instance. Bigtable participates in pipelines through Spark/Dataproc — there is currently no row-level ETL-Native executor.
Connection Fields
Source: GcpConnectorPlugin.ts (connector type GCP_BIGTABLE).
| Field | Required | Default | Notes |
|---|---|---|---|
bigtable_project_id | Yes | — | |
bigtable_instance_id | Yes | — | One connection maps to one instance. |
auth_method | Yes | — | SERVICE_ACCOUNT_KEY · WORKLOAD_IDENTITY · APPLICATION_DEFAULT. Recommended: WORKLOAD_IDENTITY. |
service_account_email | For WORKLOAD_IDENTITY | — | |
bigtable_app_profile | No | default | |
bigtable_table_id | No | — | Table to read/write. |
bigtable_column_family | No | — | Column family used for writes. |
gcp_auth_ref | No | — | Reuse credentials from another GCP connector. |
service_account_key_json (secret) | For SERVICE_ACCOUNT_KEY | — |
Required IAM Roles
roles/bigtable.reader— read-only access to the instance.roles/bigtable.user— read and write access to the instance.
Schema Behaviour
Bigtable has no fixed schema. Metadata browsing (describeTable()) surfaces column families only — individual column qualifiers only exist per-row at read time and are not part of any discoverable schema.
Spark reads and writes use com.google.cloud.spark.bigtable:spark-bigtable_2.12:0.9.1.
Known Limitations
There is no ETL-Native row-level executor for Bigtable. ETL-Native support is limited to metadata browsing and connection testing — actual row read/write only happens through the Spark/Dataproc path.