What is the best alternative to MinIO for self-hosted S3?

What did MinIO change, and when?

MinIO wound down its free edition step by step, and wrote each step down itself. The table quotes its release notes, README commits and blog.

Which self-hosted S3 servers replace MinIO?

Six projects shipped a release in October or November 2026. Each row comes from the project's own release page or documentation.

Garage or MinIO: which suits a single server?

On one server, Garage is the smaller program and MinIO the richer S3 API. Garage needs 1 GB of RAM and ships as one binary. It has no versioning, no object lock and no bucket policies; access is granted per key and per bucket.

Garage was built to spread copies across sites, not drives. Its quick start warns that a single-node layout has no redundancy. On a rented server, build RAIDZ2 first, then point Garage's data directory at the pool.

If you need object lock for backups, Garage is the wrong replacement; SeaweedFS, Ceph RGW or RustFS document it. The setup steps for Garage on ZFS are on how to host your own S3 storage on a rented server.

SeaweedFS, RustFS or Ceph: which one for growth?

All three grow past one machine; they differ in how much they ask of you.

• SeaweedFS splits a master, volume servers, a filer and the S3 gateway. weed server -s3 starts them together on one host. Its erasure coding needs no extra servers, and it protects against server failures from four servers up.
• RustFS describes itself as MinIO's simplicity rebuilt in Rust, with erasure coding and bitrot protection. It reached 1.0.0 on 16 September 2026, so its track record is the shortest here.
• Ceph RGW is the most complete and the most demanding. Its docs advise 6 GiB of RAM per hard-drive OSD, which is about 60 GiB for ten drives.

Our view: for a new single-server deployment that may grow, SeaweedFS is the safest bet today; RustFS is the one to watch. It is a judgement from the documents above, not a benchmark: we publish no speed tests.

How do I move my buckets off MinIO?

Copy bucket by bucket with an S3 client that speaks to both servers, then switch the endpoint in your apps. rclone's S3 backend names MinIO, Ceph and SeaweedFS as providers, and its "Other" setting covers the rest.

# remotes "old" (MinIO) and "new" (Garage, SeaweedFS, RustFS...) defined with: rclone config
rclone sync old:backups new:backups --transfers 16 --checksum --progress
# check before you switch the endpoint
rclone check old:backups new:backups

Where do I run the replacement?

Any machine with many drives you control works; price per terabyte decides the rest. According to Basalthold's survey of 2 November 2026, crypto hosts that ask no ID at signup charged a median $4.86 per raw TB a month at 72 TB or more. Hold 100 works out at $3.77/TB and Hold 160 at $3.50/TB.

The operating system sits on its own SSD (two mirrored SSDs on the larger holds); the data drives arrive blank and unpartitioned, so you can build ZFS, an mdadm array or a set of independent disks. If you want the team to prepare a RAID 6 or RAID 10 array before delivery, say so to support before the server is handed over. When a drive fails, open a ticket: the team has it replaced at no cost, and rebuilding the array on the new drive is up to your setup. All four holds and their drives are on the Basalthold pricing page. For locked copies, read how to make backups immutable.