Sanity tests are failing due to the lack of volume tracking. MooseFS doesn't have a ready mechanism outside of directory structure, so controller needs to track separately.
For resiliency, controller pod should mount the root directory into its own pod. Might be able to use boltdb, sqlite, csv, or a different flat file.
Additionally, the tracking could include mounted volumes, to aid in recovery. Node drivers could re-load and re-mount after restart by fetching the list of mounts from controller and the tracking system.
Sanity tests are failing due to the lack of volume tracking. MooseFS doesn't have a ready mechanism outside of directory structure, so controller needs to track separately.
For resiliency, controller pod should mount the root directory into its own pod. Might be able to use boltdb, sqlite, csv, or a different flat file.
Additionally, the tracking could include mounted volumes, to aid in recovery. Node drivers could re-load and re-mount after restart by fetching the list of mounts from controller and the tracking system.