The original intent of graceful shutdown on imminent power failure was to prevent the filesystem damage caused by sudden removal of server power, but the strategy could backfire and require a site visit for server restoral.

The S8700 executed a "shutdown -h now" command when it received the 3-minutes-to-battery-exhaustion SNMP trap from the UPS's SNMP agent. That command caused the server to exit CM and Linux software and go into the APMS firmware (part of the BIOS). The server did NOT power itself off, because the APMS firmware prevented that. If the power outage lasted long enough for the UPS batteries to run completely down and the server to completely power off, then things were fine, because the server would then automatically boot back up on restoral of power. However, if commercial power was restored before the UPS batteries gave up completely, then the server would be stuck in the APMS (BIOS) firmware and would not reboot. The only way to get the server back then would be to hit the reset button on the S8700 front bezel, or else physically disconnect and reconnect the power cable (or use an external circuit breaker to achieve the same effect). Both of those actions required a site visit, i.e., a longer service outage and more cost and inconvenience.

The CM1.X releases for the S8700 used the ReiserFS journaling filesystem, which was supposed to be highly resistant to filesystem damage during spontaneous server shutdowns, but it turned out to have some weak points, and the graceful shutdown was intended to prevent those weak points from causing filesystem damage. With CM2.0, we replaced the ReiserFS with the Linux ext3 filesystem, a journaling filesystem that has proven to be much more robust, and the graceful shutdown is no longer required, so the relevant code has been removed from the software base.