Updated October 25, 2026
· ~3 min read
For years the answer to "S3, but on our own Kubernetes" was MinIO. That changed in stages: MinIO stopped publishing community binaries and container images in October 2025, moved the community project into maintenance mode, and on April 25, 2026 archived the repository with the notice "THIS REPOSITORY IS NO LONGER MAINTAINED", pointing users to its AIStor Free and AIStor Enterprise editions. The code that remains is AGPLv3 and read-only.
So the self-hosted object storage question is open again, at the same moment S3 itself kept getting cheaper to use for new workloads — S3 Vectors, S3 Tables for Iceberg, and request pricing that rarely matters below very large scale.
This compares S3 with the realistic self-hosted options on Kubernetes: MinIO's commercial editions, Ceph through Rook, Garage and SeaweedFS.
Nothing stops working on the day a repository is archived — but there will be no security fixes, and in October 2025 the official images and binaries stopped, so every rebuild is now from source you maintain. Inventory where it runs (it is often inside other products' Helm charts as a default dependency), decide per use: on AWS, move the data to S3 and point clients at it — the S3 API is the point of MinIO, so client code rarely changes; on your own hardware, choose AIStor if you want continuity and support, or Ceph, Garage or SeaweedFS if you want an open-source licence. Test the specific S3 features you use — multipart uploads, versioning, object lock, presigned URLs, event notifications — against the replacement before moving data.
Disks are the cheap part. Durability comes from replication or erasure coding across failure domains you have to provide — separate nodes, racks or sites — and from someone who knows what to do when a disk, a node or a network partition misbehaves at 3am. Compare S3's price against the full cost: raw capacity times the replication or erasure-coding overhead, spare capacity for rebuilds, the hardware refresh cycle, and a real share of an engineer's time. For most teams already on AWS that comparison is not close.
Question 1 of 2
Can the data live on AWS?
If the data lives on AWS, use S3 — eleven nines of designed durability, no capacity planning and nothing to upgrade are hard to beat at any price you can build yourself. Self-host only when data must sit on your own hardware or in an environment without S3. For that, Ceph with Rook is the most proven at scale but asks for real storage expertise; Garage is the lightest option for small, geo-distributed clusters; SeaweedFS suits very large numbers of small files; and MinIO AIStor is the path of least change for existing MinIO users willing to take a commercial licence. Do not start anything new on unmaintained MinIO community code.
AWS Architecture Review
A focused, senior-level review of your current AWS architecture, with a prioritized list of what to fix and why.
AWS Data Lakehouse Platform
A production data lakehouse on AWS (S3-backed data lake with Glue, Iceberg, Athena, and Redshift) structured for analytics, ML feature stores, and governed data sharing.
Production EKS Platform
A production-ready Kubernetes platform on EKS with Karpenter, GitOps, service mesh, full observability stack, and security hardening: built to run real workloads.
We work with startups and growing companies on AWS architecture decisions. Tell us about your situation and we will recommend the right approach.
As of October 25, 2026. Provided as is, for general information, and accurate to the best of our knowledge on the date shown. Not legal, compliance, tax, financial or other professional advice. Prices, limits and vendor services change; check current vendor documentation before relying on anything here. Terms of use.