Hey everyone, I just realized that the MinIO repository was archived on February 13. I've spent the last two days on a hunt for a K8s-ready S3 object storage solution, but it's been frustrating. I ran into issues with Docker Hub pulls failing, broken scans, and Helm charts that seem outdated and overly complex, especially with StatefulSets. Here are some solutions I've checked out:
- **Garage**: A decentralized Helm solution, but it's tricky with single-node PV. While the LMDB backend is robust, the layout configuration is quite complex.
- **SeaweedFS**: It scales nicely, but it consumes a lot of resources. The new weed mini command does make development and testing simpler.
- **RustFS**: It's speedy for small objects but only offers basic manifests. Plus, I'm concerned about potential CLA issues.
- **Ceph**: Very reliable at scale, but it's overkill if you're not utilizing at least 1PB. Rook does help, but it still requires a dedicated team.
- **Minimus**: This seems like a solid drop-in replacement for MinIO. It has a zero CVE base and auto-patching features. I simply swapped the image tags, and everything worked seamlessly.
What are other people using for K8s-compatible S3 solutions now that MinIO is no longer an option? Looking forward to hearing your experiences!
5 Answers
From my experience, Ceph remains the most stable option from your list. With Rook, it’s pretty straightforward to deploy. People say you need a whole team, but I honestly found it no more complex than MinIO – think they must not have fully tested it.
Right? I had to migrate from MinIO to Ceph, and the transition was quite manageable.
I think CNCF should step up and support initiatives that help maintain alternatives to MinIO. A lot of companies rely on it, and there should be resources for developers to keep these projects alive.
I'd love to see that happen, but CNCF has strict rules against paying devs, which is kind of a bummer.
Two days of testing really isn't enough time to draw strong conclusions about those solutions. What kind of tests did you conduct, and what scale are you working with? Just curious about your testing methods and the actual results you got! Seems a bit hasty to dismiss without deep dives.
That's true, seems pretty bold to come to those conclusions. I've had my own challenges with Ceph but with proper setup, maintenance has been less than I expected.
Yeah, what level of qualifications do you have to back those test claims? Just a thought.
Has anyone considered what other services like GitLab and Harbor will do now that MinIO is archived? I switched to RustFS after the news broke, and so far, it’s been solid even in a homelab setup.
That sounds promising! I feel like most services will just adapt by requiring an S3 provider upfront during setup.
I recently learned about a new option called S4, which uses the Ceph RADOS gateway and aligns well with standard file systems. Definitely check it out if you’re looking for alternatives!
Exactly my point. Setting up a CephCluster takes just a few minutes, and then adding other services is smooth.