Object Store Options
Several connectors read or write cloud object storage through an object store client. Keys that start with obstore. configure that client. Set them where the connector takes its other settings: conn_cfg on the connection for the file connectors, and cfg on the staging location for ClickHouse, StarRocks and Druid.
Where the keys apply
| Connector | Set with | What the keys configure |
|---|---|---|
| Amazon S3, Google Cloud Storage, Azure Blob Storage | conn_cfg on the connection | Reads and writes |
| ClickHouse | cfg on the staging location | The staging upload, the address ClickHouse reads the files from, and the cleanup |
| StarRocks | cfg on the staging location | The staging upload. StarRocks reads the files through its own path, which the keys do not change |
| Druid | cfg on the staging location | The staging upload. The ingestion input source does not use them |
On the file connectors, a key in conn_cfg overrides the matching credential in the connection. The explorer ignores these keys.
Keys by provider
| Provider | Keys |
|---|---|
| Amazon S3 | access_key_id, secret_access_key, session_token, region, endpoint, virtual_hosted_style_request, allow_http, anonymous, skip_signature |
| Azure Blob Storage | account_name, account_key, sas_token, endpoint, allow_http |
| Google Cloud Storage | service_account_key, service_account, allow_http |
service_account is the path to a key file on the Tabsdata server. Raw option names with the aws_, azure_ or google_ prefix are accepted as well.
For example, to point an S3 connection at a MinIO server:
conn_cfg:
obstore.endpoint: 'http://minio.internal:9000'
On ClickHouse, an unknown obstore.* key fails the write.