MinIO's GitHub repository was archived in April 2026. The code is still there under AGPLv3, but nobody upstream is patching it, and the community build had already lost its admin console in 2025. If you run MinIO today, you are running object storage that will not get another security fix.
This is a survey of the open source S3-compatible stores teams are moving to, what each one is built for, and where each one falls short. One of them is ours, and we say where it falls short too.
The short answer
RustFS is the closest drop-in for a MinIO cluster: same design, Apache-2.0, and a 1.0 release in September 2026. Garage suits small clusters spread across sites. SeaweedFS handles billions of small files. Ceph is the most complete and the heaviest to run. Predastore fits when you want object storage that shares IAM with an AWS-compatible cloud on your own hardware.
| Project | Licence | Language | Built for | Watch out for |
|---|---|---|---|---|
| RustFS | Apache-2.0 | Rust | Replacing MinIO directly | 1.0 shipped in September 2026, so its production record is short |
| Garage | AGPL-3.0 | Rust | Small, geo-distributed self-hosted clusters | Replicates rather than erasure-codes, and implements a subset of S3 |
| SeaweedFS | Apache-2.0 | Go | Billions of small files, plus a filesystem view | More moving parts than a single-binary object store |
| Ceph (RGW) | LGPL | C++ | Object, block and file storage at scale | Operational weight: plan for a team that knows Ceph |
| Versity Gateway | Apache-2.0 | Go | An S3 API in front of an existing filesystem | Durability comes from the filesystem underneath, not the gateway |
| Predastore | AGPL-3.0 | Go | S3 that shares IAM with EC2 and EKS under Spinifex | No lifecycle policies, object lock or bucket policies yet |
What happened to MinIO
MinIO spent 2025 moving its engineering effort to AIStor, its commercial
product. The community build lost its admin console, and the project stopped
publishing community binaries. In April 2026 the
minio/minio repository was archived and made read-only.
Existing deployments keep working. What stops is the flow of fixes, and for a service that sits on the network holding your data, that is the part that matters. Forks exist, but none has emerged as the agreed successor, which is why most teams are choosing a different project rather than a fork.
RustFS
RustFS is the project built most deliberately as a MinIO replacement. It follows MinIO's model of erasure-coded storage across a pool of drives, ships a web console, and is designed to run alongside an existing MinIO or Ceph cluster during a migration. It is Apache-2.0, which removes the AGPL question for teams whose lawyers ask it.
The trade-off is age. RustFS reached 1.0 in September 2026. It has a large and fast-growing community, but far less time in production than MinIO had, so test your own failure cases before you trust it with the only copy of anything.
Garage
Garage comes from Deuxfleurs, a French collective that runs its own infrastructure across several homes and small sites. It is built for exactly that: a handful of modest machines in different locations, connected over ordinary internet links, staying available when one site drops out.
Garage replicates objects rather than erasure-coding them, which costs more raw disk and buys simplicity. It also implements a deliberate subset of S3, so check its compatibility list against the API calls your applications make. For a homelab or a small multi-site deployment, it is the lightest serious option on this list.
SeaweedFS
SeaweedFS started as an implementation of Facebook's Haystack design, and its strength is still huge numbers of small files with constant-time disk access. On top of the blob store it offers an S3 gateway, a POSIX-style filer, and Iceberg table support.
That breadth means more components to deploy and understand than a single-binary object store. If your workload is millions of thumbnails, log chunks or ML training samples, it is worth that cost.
Ceph
Ceph's RADOS Gateway is the
most complete open source implementation of the S3 API, to the point that
the standard S3 conformance suite, ceph/s3-tests, comes from the
Ceph project. Ceph also gives you block and file storage from the same
cluster.
The cost is operational. Ceph rewards teams that already know it and punishes teams that do not, and a MinIO deployment of four nodes is rarely the right place to start learning.
Versity Gateway
Versity Gateway takes a different approach: it translates S3 calls onto a filesystem you already run, such as ZFS, a POSIX mount or a parallel filesystem. If your data already lives on well-managed storage and you only need an S3 API in front of it, this is the shortest path.
Predastore
Predastore is our object store. It erasure-codes objects with Reed-Solomon across blob nodes, keeps its object index in a Raft quorum, moves data between hosts over QUIC, and encrypts data at rest with AES-256-GCM. It runs standalone or as the object storage of Spinifex, our open source AWS-compatible cloud.
Standalone, Predastore authenticates with access keys from its config file, and every key can reach every bucket. That suits a single application or a single team. Under Spinifex, the same access keys, IAM users, roles and policies that govern EC2 and EKS govern bucket access, and an EC2 instance with an instance profile reads its bucket with no credentials configured in the guest.
Predastore does not yet implement lifecycle policies, object lock, bucket
policies, ACLs or CORS, and shard rebalancing is manual. We publish its
results against ceph/s3-tests in the
repository. If you
depend on any of those features today, pick another store on this list.
Moving off MinIO
Every option here speaks the S3 API, so the migration has the same shape:
stand up the new cluster, recreate users and access policies, copy buckets
with rclone, then repoint clients. We wrote a step-by-step
version for Predastore in
Migrating from MinIO to Predastore,
and the rclone half of it applies to any target.
To try Predastore and Spinifex before deploying on your own hardware, sign up for the free sandbox, or read the documentation at docs.mulgadc.com.
Frequently asked questions
Is MinIO still maintained?
No. The minio/minio GitHub repository was archived in April 2026 and is read-only. The AGPLv3 code is still available, but the community edition gets no further features, bug fixes or security patches. MinIO's engineering now goes into AIStor, its commercial product.
What is the best open source alternative to MinIO?
RustFS is the closest drop-in replacement for a MinIO cluster: erasure-coded, Apache-2.0, with a 1.0 release in September 2026. Garage suits small geo-distributed clusters, SeaweedFS suits billions of small files, Ceph suits teams that need object, block and file storage together, and Predastore suits teams running an AWS-compatible cloud with Spinifex.
RustFS or Garage: which should I use?
RustFS if you are replacing a MinIO cluster and want the same erasure-coded design under Apache-2.0. Garage if you run a few modest machines across different sites and value simplicity over storage efficiency, since it replicates rather than erasure-codes.
Can I still use MinIO?
Existing deployments keep running, but they will not receive security fixes. For anything exposed to a network or holding data you cannot lose, plan a migration to a maintained S3-compatible store.
How do I migrate from MinIO?
Stand up the new cluster, recreate users and access policies, copy each bucket with rclone, then repoint clients at the new endpoint. Every option in this comparison speaks the S3 API, so applications only need an endpoint and credentials change.
