MinIO

S3-compatible object storage. The data lake layer — lands raw files alongside the warehouse, stays reachable for ML and DS consumers.

Tier

Cold. MinIO is lightweight (~200 MB) but classified cold because the typical consumption pattern is bursty (ad-hoc dataset pulls into notebooks, periodic ML training reads) rather than always-on. The orchestrator brings it up on first read access and stops it after the idle timeout. If you have a workload that needs MinIO available 24/7, pin it from the dashboard's Keep warm card.

Sign-in: MinIO's web console runs its own login. From the dashboard, open MinIO's Details · pin · activity page → in the Connection card click Reveal to surface the root user + password the wizard captured. Admin-only; the same credentials work as S3-style access key / secret for boto3 + DuckDB + Polars clients.

Role

MinIO is the "open end" of OrcheStack. Airbyte writes every sync to MinIO in parallel with PostgreSQL. The warehouse track (Airbyte → Postgres → dbt → Metabase) serves BI; the lake track (Airbyte → MinIO → external consumers) serves ML/DS use cases.

MinIO is never upstream of dbt. dbt only reads from PostgreSQL.

Bucket layout

OrcheStack/
├── raw/                  # Airbyte-landed files, partitioned by date
│   └── <source>/
│       └── <YYYY-MM-DD>/
├── backups/              # nightly dumps of PostgreSQL
└── exports/              # user-triggered data exports from the dashboard

Configuration

• MINIO_ROOT_USER — administrator access key
• MINIO_ROOT_PASSWORD — administrator secret key
• MINIO_BUCKET — top-level bucket name (default: OrcheStack)

Accessing externally

MinIO exposes the standard S3 API on :9000. Point any S3 client at http://<your-host>:9000 with the root credentials, or generate service accounts from MinIO's web UI at /app/minio.