Stability on NFS and/or RAID backing storage

[omarsamir27]

I used to run a Minio cluster on Kubernetes with the backing enterprise storage that is configured in RAID and connected to Kubernetes via NFS PVC.
The cluster started being unstable with 50% down time once the files started increasing past a certain point.
Using NFS/RAID with object storage with erasure coding is advised against according to Minio Blog
https://blog.min.io/erasure-coding-vs-raid/
https://blog.min.io/no-san-nas/
After I migrated to Minio DirectPV option, the cluster is stable.
I know ask what the recommended hardware setup for RustFS is, does it have the same needs because it also implements erasure coding?

[loverustfs]

Hi @omarsamir27 ,

Yes, you can imagine that we are the Rust version of MinIO. Our requirements are exactly the same as MinIO.
We also do not recommend using NFS for k8s disk mounting.

Using NFS for disk mounting is OK in single-machine single-disk mode, but it will cause a lot of trouble after multiple servers and multiple disks.

So, you can use DirectPV, LongHorn, OpenEBS, etc. as block storage to provide to RustFS.