FAQ

Is this S3-compatible today?

No. It is S3-inspired and currently in MVP build stage. The focus today is reliable core behavior before broad API parity.

Does it support multipart upload?

Not yet. Planned for post-MVP expansion.

Is replication durable?

Yes, replication jobs are persisted in PostgreSQL and retried on failure.

How are duplicate replica writes handled?

Idempotency is enforced through DB constraints and conflict-safe writes.

What consistency model does download use?

Download resolves latest version from metadata (objects.is_latest = TRUE) and serves that version. Secondary-node replication is asynchronous, so replica convergence is eventual.

Why not replicate synchronously before upload response?

That would increase client latency and couple availability to every replica node. Current design acknowledges upload after durable primary + metadata + replication intent, then processes replicas in background.