During an upgrade from 4.11 to 4.12 of an AWS IPI cluster with prometheus storage allocated using in-tree drivers experienced an unexpeted behavior was observed. 
1.- KCM detach the volume. (in-tree)
2.- Node becomes restarted.
3.- After restart, EBS CSI driver start reporting that the volume was not attached into the node.
4.- Node description says that volume is mounted and in use

4.12.17

1.
2.
3.

This was fixed by attaching the prometheus volume to the instance manually from AWS side.