The current approach, letting the CSI Controller generate the ConfigMap entries for DRBD Reactor has some drawbacks:
- It seems easy to "de-synchronize" the state of LINSTOR volumes with what is in the ConfigMap.
- The ConfigMap requires special handling in deployments, as it gets created there, but needs to be updated by the CSI driver.
An alternative would be to let the nfs-helper tools generate the configuration based on either:
- The PV information
- The LINSTOR RD properties
Both of these are already under direct control of LINSTOR CSI, so it should be much harder for them to get in weird states.
The current approach, letting the CSI Controller generate the ConfigMap entries for DRBD Reactor has some drawbacks:
An alternative would be to let the nfs-helper tools generate the configuration based on either:
Both of these are already under direct control of LINSTOR CSI, so it should be much harder for them to get in weird states.