MinIO's Community Edition Is Archived: Garage vs. SeaweedFS vs. RustFS Compared

MinIO's Community Edition GitHub repository is archived. Read-only. No new commits, no reviewed patches, no official binaries — the culmination of an 18-month retreat that started with pulling the web console in May 2025, continued with cutting off pre-built Docker images and binaries in October 2025, and ended with the repo marked "no longer maintained" on February 12, 2026 before being locked for good on April 25, 2026. If your self-hosted PaaS's storage-class default still says minio, here's the concrete replacement comparison that decision needs:

That table is the whole recompute in miniature: three genuinely different projects, with three genuinely different risk profiles, and the risk that matters most isn't RAM — it's whether the governance model underneath the one you pick can pull the same move MinIO just did.

What Actually Happened to MinIO​

MinIO's retreat from open source didn't happen in one announcement — it happened in three deliberate steps, each one a little further than the last.

Step one, May 2025: MinIO stripped the management console — the web GUI for bucket management, IAM policy editing, and audit logging — out of the Community Edition entirely, via a commit filed by co-founder Harshavardhana with no changelog warning and no advance notice. According to Blocks & Files, users discovered the change only after upgrading.

The console kept working — but only in MinIO's paid AIStor product. Community Edition users were pushed to the mc CLI for administration that used to be a few clicks in a browser. The backlash was immediate: a Hacker News thread, a wave of GitHub issues, and a discussion thread on the MinIO repo itself titled "It's not a feature issue, it's a trust one".

Step two, October 2025: MinIO stopped publishing Docker images and pre-built binaries for the Community Edition. From that point forward, running it meant compiling from source — a small but deliberate friction increase for anyone who'd built an automated deploy pipeline around docker pull minio/minio.

Step three, February–April 2026: MinIO flipped the repository's status from "maintenance mode" to "no longer maintained" on February 12, 2026, archived it shortly after, briefly unarchived it, then locked it permanently on April 25, 2026, according to The Cloud Support Engineer and Storm Developments. No new releases. No reviewed patches. No guaranteed security fixes, ever again, for the free tier.

What's left for teams who want the console, ongoing patches, and official support is MinIO AIStor, MinIO's commercial product, which now starts at roughly $24,000/year and scales well past six figures for petabyte-class deployments. Existing Community Edition installs keep running — MinIO isn't deleting anyone's data — but they're now frozen software: whatever CVEs surface against the last released version stay unpatched unless you're paying for AIStor or backporting fixes yourself.

The pattern is familiar. HashiCorp relicensed Terraform away from MPL in 2023. Redis moved off the BSD-style license in 2024. Each time, the open-source tier didn't disappear outright — it got starved: features frozen, binaries withheld, support redirected to the paid product, until "still technically open source" and "actually usable in production" stopped meaning the same thing. MinIO followed the same script, just spread across three separate moves instead of one relicensing announcement.

The Three Replacements, Compared​

Garage is the smallest and most deliberately scoped of the three. Built by Deuxfleurs, a French nonprofit collective, it's a single Rust binary with zero external dependencies — no ZooKeeper, no etcd, no separate metadata store. Its own docs specify it runs on as little as 1GB of RAM and 16GB of disk, including on ARM boards like a Raspberry Pi, with an internal object cache that defaults to just 128MB.

Garage was designed from day one for small, geographically distributed self-hosted deployments — think a handful of nodes spread across a home lab, a few VPS providers, and a Hetzner box, not a single datacenter's worth of racks. The tradeoff for that footprint: Garage's S3 API surface is intentionally narrower than MinIO's or AWS's — no bucket lifecycle policies, no object versioning as of the current stable release. For a self-hosted PaaS's tenant-facing bucket storage — app uploads, build artifacts, backups — that's rarely a blocker. For a team that specifically needs S3 lifecycle rules today, it is.

SeaweedFS is the opposite bet: maximum scale, maximum maturity. Created by Chris Lu (ex-Uber) and now sitting at 28.2k GitHub stars, 2.6k forks, and over 10,000 commits, it's designed to handle billions of files with O(1) disk seeks and supports S3, POSIX FUSE mounts, WebDAV, Hadoop, and cross-datacenter replication out of the box.

The strongest external validation it has is architectural, not marketing: as of Kubeflow Pipelines 2.15, SeaweedFS replaced MinIO as KFP's default bundled object store — a CNCF-adjacent ML platform choosing SeaweedFS specifically because it needed a default that wasn't tied to a single company's commercial roadmap. Its governance is still effectively one lead maintainer plus a large contributor base, not a foundation — a real single-point-of-continuity risk, if a smaller one than MinIO's commercially-directed roadmap posed. And its operational cost is complexity: a full SeaweedFS deployment runs separate master, volume, and filer components, a heavier operational surface than Garage's single binary.

RustFS is the newest and the fastest-growing by star count — over 28,000 stars in roughly ten months, an Apache 2.0 license (explicitly marketed as avoiding AGPL's copyleft terms), and vendor-published benchmarks claiming 2.3x MinIO's throughput on 4KB object payloads. It joined NVIDIA's Inception accelerator program in April 2026 and is positioning hard toward AI/data-lake workloads with native S3 Tables support.

Here's the catch worth sitting with: RustFS is trademarked to RustFS, Inc., and its own site runs an "Invest in RustFS" section describing a coming commercial tier with paid support, SLAs, and a funded engineering team — the same starting shape MinIO Inc. had years before it started narrowing its free tier. Apache 2.0 means RustFS can't repeat MinIO's exact move — you can't retroactively strip features from code already under a permissive license and call it enforcement — but a corporate entity funding development through a future paid product is a governance signal worth tracking, not dismissing because the license today is friendlier.

Governance Is the Metric That Actually Predicts What Happens Next​

RAM footprint and star count are the numbers every comparison post leads with, and they matter for capacity planning. But MinIO wasn't AGPL-licensed open source that suddenly stopped working — the code from before May 2025 still runs today, unpatched. What changed was who controlled the roadmap and what business model that control served. That's the axis a self-hosted PaaS's storage-class default actually needs to be picked against, because it's the one that predicts whether you're doing this recompute again in eighteen months.

Ranked on governance risk, not raw capability:

• Garage carries the least commercial pressure of the three. Deuxfleurs is a nonprofit with no paid tier to push Community Edition users toward — there's no AIStor-shaped upgrade path sitting behind Garage's free version, because there's no revenue model requiring one.
• SeaweedFS sits in the middle. No named company monetizes it today, and Kubeflow's adoption as a default is a real, external vote of confidence in its independence — but a project this dependent on one lead maintainer's continued involvement is a bus-factor risk MinIO (a funded company with a payroll) never actually had, even as its commercial incentives cut the other way.
• RustFS carries the license protection Garage and SeaweedFS don't strictly need (Apache 2.0 can't be un-given the way MinIO stripped features under AGPL) but the most visible early-stage commercial intent — a company literally soliciting investment around the same codebase you'd be running for free.

None of that makes RustFS a bad choice today — the performance numbers and NVIDIA backing are real, and Apache 2.0 is a structurally stronger legal floor than what MinIO ran on. It means "which one will still be fully-featured and unpatched-free in three years" doesn't have the same answer as "which one benchmarks fastest this quarter," and a platform's default shouldn't conflate the two.

What a Self-Hosted PaaS Should Actually Default To​

Match the store to the workload, not to whichever benchmark result is loudest:

For bex specifically, the takeaway is concrete: docs and tenant defaults that still point at MinIO are pointing at a project whose free tier is now frozen software, not a live recommendation. Garage's footprint and governance make it the better default for the common case — a tenant's app storage bucket — precisely because a self-hosted PaaS's storage-class recommendation shouldn't need revisiting every time the vendor behind it decides its open-source tier is worth less than its enterprise one.

Bex.co is the open-source, AI-native Render alternative — push a git repo, get a running HTTPS service on machines you own, with storage defaults that don't depend on a vendor's enterprise upsell staying out of your way. Star the repo on GitHub or deploy your first app today.

Sources:

• The End of an Era: MinIO Community Edition is Archived – What's Next?
• MinIO's community edition is archived. What still runs in 2026?
• MinIO users complain after admin UI removed from Community Edition — Blocks & Files
• "It's not a feature issue, it's a trust one" — MinIO GitHub Discussion #21326
• MinIO AIStor Pricing
• Garage — Deuxfleurs
• Garage GitHub — deuxfleurs-org/garage
• SeaweedFS GitHub — seaweedfs/seaweedfs
• Object Store Configuration — Kubeflow Pipelines Docs
• RustFS GitHub — rustfs/rustfs
• RustFS — The High-Performance, Apache 2.0 Alternative to MinIO