Currently, installing pre-provisioned (deployed servers) OSP17.1.4 as a greenfield, and also updating it from OSP17.x, require a compute node reboot in order to apply the delayed installation of the systemd-container package. See the background story for "why it does" in bz2247302.

This situation is OK for the OSP17 lifecycle. However, it became a blocker for the adoption of workloads from OSP17 to RHOSO18. We should not allow adoption of OSP17, when there is no systemd-container package installed on all compute hosts assigned to node sets.