FS permissions causes long rollout times

[moray95]

Describe the bug Running RustFS in Kubernetes using the official chart causes permission fixes on the file system twice:

• Once by Kubelet because of fsGroup specified under securityContext
• Once by the init-step container running chown

The total time for a RustFS pod to become up and running is in the range of 15-20 minutes in our environment because of this.

To Reproduce

• Deploy RustFS using the Helm chart
• Write some data. Our current setup has around ~600k files per drive (counted using find . | wc -l).
• Start a rollout of the RustFS StatefulSet

Expected behavior Each pod becomes available in a reasonable timeframe (~2-3 minutes).

Additional context Running a cluster on Ubuntu 24.04, RKE2 1.33.5.

I wanted to ask if the permission fixes are necessary at all? If necessary, is it really necessary to have two different mechanisms for this? With more files, these operations will most likely take more and more time, which does not seem to be viable long-term.