MinIO’s open-source server is done. The GitHub repository was archived on April 25, 2026, after a year of features being pulled out of the community edition. Your existing MinIO cluster still runs, but nobody is fixing it anymore, and the next CVE won’t get a community release. If MinIO holds anything you care about, it’s time to pick where that data goes next.
Here’s what happened, the open-source S3 servers worth moving to, and which ones fit which job. Every version number and feature below was checked against each project’s own site or GitHub on October 16, 2026.
What happened to MinIO
MinIO moved its server from Apache 2.0 to the AGPLv3 license in 2021. The real changes came later:
- 2025: the admin features were stripped out of the community console, which was reported in June 2025. Prebuilt Docker images reportedly stopped that October.
- December 3, 2025: the community edition went into maintenance mode, with no new features, no pull request reviews and no more packages, according to InfoQ.
- February 2026: the README started saying the project is no longer maintained.
- April 25, 2026: the repository was archived.
MinIO’s product is now AIStor. AIStor Free runs on a single node under a proprietary license, and distributed clusters need AIStor Enterprise. That’s a fine choice if you want to pay MinIO. If you picked MinIO because it was open source, it isn’t a MinIO you can use anymore.
The alternatives at a glance
| Server | License | Object Lock | Versioning | Data protection | Smallest sensible setup | Latest release |
|---|---|---|---|---|---|---|
| Ceph RADOS Gateway | LGPL | Yes | Yes | Replication or erasure coding | About 3 nodes | Tentacle 20.2.4 (Aug 2026) |
| SeaweedFS | Apache 2.0 | Yes | Yes | Replication; erasure coding for warm data | 1 node | 4.48 (Sep 2026) |
| Garage | AGPLv3 | No | No | Replication only | 3 nodes | 2.4.1 (Sep 2026) |
| RustFS | Apache 2.0 | Yes | Yes | Erasure coding | 1 node | 1.0.1 (Oct 2026) |
| Silo (MinIO fork) | AGPLv3 | Yes, as MinIO | Yes | Erasure coding | As MinIO | Sep 2026 release |
| Versity S3 Gateway | Apache 2.0 | Yes | Yes (POSIX backend) | None of its own | 1 node | 1.9.0 (Oct 2026) |
| Apache Ozone | Apache 2.0 | No | No | Replication or erasure coding | Several nodes | 2.2.0 (2026) |
If the data is backups, read the Object Lock column first. Object Lock is what makes a backup immutable, and backup tools like Veeam need it. That rules out Garage and Ozone for that job, however good they are at everything else.
Ceph RADOS Gateway: the safe bet at scale
Ceph’s S3 gateway has been in production for over a decade, and it does the S3 features people actually miss after MinIO: Object Lock in both governance and compliance modes, versioning, lifecycle rules and multisite replication. The current release line is Tentacle (20.2.x). If you’re still on Squid (19.2.x), its estimated end of life is October 31, 2026, so plan the upgrade with the migration.
The catch is Ceph itself. It’s the most work to run of anything on this list. You’ll want three nodes at minimum so it can keep three copies on three machines, and someone on the team who’s comfortable with placement groups and slow-disk alerts. Get past that and it scales further than anything else here.
SeaweedFS: the closest thing to a MinIO swap
If you liked MinIO because it was one binary you could start on one box, SeaweedFS is the nearest match. It’s Apache 2.0, written in Go, and runs fine on a single node. It also supports Object Lock with governance, compliance and legal hold, which puts it on the short list for backups. Hot data is replicated, and warm data can be erasure coded.
The trade-off: some features sit in its paid enterprise edition, including custom erasure coding layouts. Check that the community edition covers what you need before you commit.
Garage: great for small clusters, wrong for backups
Garage gets a lot of search traffic, and it deserves it for what it’s built for: small, geo-distributed clusters on modest hardware. It’s AGPLv3, written in Rust, and its v2 line is shipping steadily. It replicates data three ways across nodes rather than using erasure coding.
It doesn’t do versioning or Object Lock. Garage’s own S3 compatibility table lists both as missing. For app uploads, static assets or a homelab, that’s fine. For anything that has to survive ransomware, it isn’t.
RustFS: young, but past 1.0
RustFS set out to be a MinIO-compatible server under Apache 2.0, and it reached version 1.0 in September 2026. It documents Object Lock with governance, compliance and legal hold, plus versioning. Its README still marked distributed mode as “under testing” when we checked. Single-node is the safer way to try it today.
Staying on MinIO code: the Silo fork
If migrating isn’t possible yet, the Pigsty team’s fork keeps the MinIO code alive under AGPLv3. It restores the console, ships packages, and was renamed Silo in August 2026, with its latest security release in September. Treat it as a bridge rather than a destination. It’s maintained by a small team, and a cluster can’t mix MinIO and Silo nodes, so all nodes move together.
The others, briefly
Versity’s S3 Gateway puts an S3 API in front of a filesystem or another object store and enforces Object Lock, but durability comes from whatever sits underneath it. Apache Ozone is a Hadoop-style system with erasure coding, but its S3 gateway lists versioning and Object Lock as not supported. Scality’s Zenko CloudServer is still maintained, though it’s more of a single-process gateway than a storage cluster.
How to pick
- Backups that must be immutable: Ceph RGW if you have three or more nodes and the skills, SeaweedFS if you want something lighter.
- App data on one or two servers: SeaweedFS, or RustFS if you’re comfortable on a 1.0 release.
- Small clusters spread across sites, no WORM needed: Garage.
- Buying time: the Silo fork, with a migration date on the calendar.
Moving the data is the easy part. rclone copies bucket to bucket over the S3 API, so you can run the new server alongside MinIO, sync, cut over and keep MinIO read-only until you’ve checked object counts. The slow part is everything that points at the old endpoint.
Where to run it
All of these want the same thing: big drives, a fast private network between nodes and a predictable bandwidth bill. Our storage servers take 14 TB enterprise drives at $79 a month each, and a 10 Gbps private network link is $20 a month per server, so replication traffic stays off your public port.
Rather not run a storage cluster at all? Our S3-compatible object storage gives you an S3 endpoint we run, in Ashburn, Chicago or Los Angeles. It has Object Lock for immutable backups, and it sits in the same building as your servers.
FAQ
Is MinIO still safe to use?
Existing clusters keep running, but the open-source edition gets no maintenance and the repository is archived. New security problems won’t be fixed in a community release, so plan a move or a supported option.
What’s the closest drop-in replacement for MinIO?
For a single-binary, single-node feel, SeaweedFS. To keep the same code for now, the Silo fork. Your apps only talk the S3 API, so most of them won’t notice the swap beyond a new endpoint and keys.
Which MinIO alternatives support Object Lock?
Ceph RGW, SeaweedFS, RustFS and the Versity gateway document it. Garage and Apache Ozone don’t support it.
Can GigeNET run S3 storage for me?
Yes. Our S3-compatible object storage is run by our engineers and quoted per deployment, starting at 10 TB. Or rent storage servers and run any of the options above yourself.
