🤔 Problem

Currently csi rclone has several shortcomings:

• multitenancy issues - especially security risks but also performance
• restarting of the daemonset loses the mounts
• we have to maintain a whole helm chart and project - csi-rclone
• security concerns and operations - people have to run a whole csi driver
• installing csi rclone on openshift or more enterprise-like environments is a real pain

🍴 Appetite

6 weeks

🎯 Solution

Option A: Mount in a sidecar inside the session pod

This requires elevated permissions on the sidecar container so that the mounts propagate to the session container.

We can have a custom process or tool that:

1. mounts all required data connectors and then
2. provides a unified health check that will fail if either the mount of the actually process fails
3. ensures the mounts are running and are active
4. checks for and maintains tokens / credentials for storage - especially the ones that may need refreshing and are not or cannot be handled by rclone natively

Option B: Mount inside the session container