Skip to main content

0016 — Adopt SeaweedFS using embedded filer S3

Proceed with SeaweedFS through the embedded FilerSpec.S3 field rather than wait for standalone S3 support.

Public edition condensed on 2026-10-21; original decision date and meaning retained.

Superseded

Replaced by ADR 0019. Read it before applying this decision.

Context​

ADR 0006 had waited for a SeaweedFS chart that exposed standalone S3 configuration. The release date looked uncertain, and Harbor needed an object backend.

Decision​

Proceed with the SeaweedFS chart then available, using the embedded FilerSpec.S3 path. Keep SeaweedFS as the in-cluster object store on Longhorn volumes, with SeaweedFS replication disabled because Longhorn provides block durability. Harbor is the named consumer. Filer high availability remained an open test requirement.

Alternatives considered​

  • Keep waiting for standalone S3: an indefinite delay with a consumer waiting.

Consequences​

The decision accepted a deprecated field and a future migration cost to avoid waiting. It was superseded by ADR 0019 before installation, when the canonical spec.s3 path shipped, so its chart and field choices are not deployment guidance. The filer HA question stayed open after the interface change.