Unix Tip: Reboot mystery solved

The system was rebooting every other Thursday without leaving any evidence except the notices in the events monitor stating that the reboot that had occurred minutes before was unexpected. This problem had been occurring for months by the time we realized it was occurring regularly. Ironically, the system was rebooting in the middle of my staff meetings -- until we moved the clocks back an hour on November 4th. I scheduled a reminder so that my blackberry would tell me to log in just prior to the next expected occurrence of the reboot, but all I was able to determine was that everything looked normal until the system froze. The processes looked normal, performance was great and no error messages or warnings were in evidence. A few minutes later, the system was back up with the usual notice about the unexpected reboot.

The first thing I do any time a strange problem occurs is enter a few search terms in Google looking for other people who have seen and, hopefully, solved the same problem. This time, however, I was getting nothing and, frankly, terms like "Windows" and "reboot" are going to generate more hits than I'd have time to review even if I were to spend the rest of my life reviewing them. Of course, I entered "two weeks" and "2 weeks", but I wasn't getting anywhere. It struck me as very strange that anything on this server that I manage would be causing such regular reboots. What, after all, would be both so regular and so badly behaved as to unceremoniously crash the system every two weeks? I know enough about Windows to look for scheduled jobs that might have developed an evil second nature and enough about the applications on this particular server to know how they work. Nothing seemed to fit the bill. It was as if my serious and unusually considerate boss threw a cup of coffee across the room -- and at a precise two-week interval.

The answer slowly came to the surface when we searched on "14 days" instead of "two weeks". First, we found this comment that had been posted years earlier by someone else who'd had a Windows system that was similarly rebooting every two weeks, followed by a response to check the UPS if one was involved:

"I have a Dell PowerEdge 2650 running W2K server SP4 which has been rebooting itself for several months now. This always occurs every 14 days, and always at the same time of day. This has not been a big issue for the users, but I would like to clear it up before it becomes one."

A little more digging and we learned that (some) APC UPS devices perform a self-test every 14 days. We checked our system. Yes, it was plugged into an APC UPS and, further, it seemed that the device's battery was going bad. We scheduled a time to move the system off the small UPS and plug it into lab power (protected by a room-sized UPS device). Problem resolved.