Background

In order to support Node Graceful Shutdown properly, MCO will need to respect systemd reboot inhibitor locks.

Scenarios

A privileged pod on a system is flashing device firmware or reconfiguring an FPGA device can be a quick operation or a 30+ minute event. In either case, rebooting the system during the operation could be catastrophic for the device.

Even with standard workloads, inhibiting reboot until pods have properly drained is desirable and required for Node Graceful Shutdown to be meaningful.