S3 and S3-compatible object storage
Read data files directly from S3-compatible object storage — AWS S3, Alibaba Cloud OSS, Tencent COS, Huawei OBS, Cloudflare R2, and MinIO. Files are read in place and can be joined with datasets from other connectors.
When to use
- Analytical files (Parquet, CSV, JSON) exported to a bucket.
- A folder of partitioned files that should be read together as one table.
Supported file formats
- Parquet, CSV, and JSON are read directly.
- A glob pattern can match multiple files so a partitioned export is analyzed as a single table, e.g.
exports/2026/*.parquet.
Before you begin
Create credentials (Access Key ID and Secret) with read-only access scoped to the bucket or prefix that holds the data. For MinIO and self-managed deployments, note the endpoint host and whether it serves plain HTTP.
Connection settings
Provide the object-storage access credentials: choose a cloud vendor preset (AWS, Alibaba Cloud OSS, Tencent COS, Huawei OBS, Cloudflare R2, MinIO, or Other), which fills the endpoint, region, URL style, and SSL setting; then supply the access keys and:
| Field | Required | Description |
|---|---|---|
| Bucket | Yes | The bucket that holds the files. |
The object path or glob for the data is specified when the dataset is defined.
A note on addressing and SSL
- URL style —
vhostputs the bucket in the hostname (most public providers);pathputs it in the path (required by MinIO and Cloudflare R2). The preset sets this correctly; change it only for a custom deployment. - Enable SSL — on for public providers. MinIO is frequently plain HTTP; turn SSL off only for such a deployment.
Test and save
Enter the fields and use Test connection. On success, define the object path or glob to read.
Capabilities
- Execution model: Federated. Files can be joined with datasets from other connectors.
- Access: Read-only.
- Vendors: Any S3-compatible provider works via the preset system; there is no separate connector per vendor. For Alibaba Cloud OSS specifically, use this connector with the "Alibaba Cloud OSS" preset.
Troubleshooting
- Access denied — verify the keys have read access to the bucket and prefix, and that Region matches the bucket's region.
- Endpoint unreachable — check the endpoint host; for MinIO confirm the port and the HTTP/HTTPS setting.
- A vendor requires path-style — MinIO and Cloudflare R2 need URL style set to
path; the preset does this automatically.