Will shutting down really help?

greenspun.com : LUSENET : TimeBomb 2000 (Y2000) : One Thread

I keep on reading where companies and utilities will be shutting down for a few hours around midnite 1/00. I can understand the general logic behind this, but how can this really help? Could one really be able to pinpoint just where their problems may be should they do so? Thanks

-- Rich (rubeliever@webtv.net), December 24, 1999

Answers

Utilities will not shut down, businesses usually shut down on New Years anyway, it is a holiday every year. No big deal. Some "IT" businesses may shut down but major businesses that usually shut down will anyway and those who do not normally shut down will let the utilities know ahead of time if they use any noticable amout of power anyway.

-- Cherri (sams@brigadoon.com), December 24, 1999.

Shutting down is a problem in itself. Anyone who ever worked in a BIG shop that has in excess of a dozen super mini or mainframe systems knows how often just cycling power on all the hardware will almost always end up killing at least one machine.

I worked in a shop that had 55 super minis, each was 3-5 boxes the size of refrigerators. EVERY SINGLE TIME we did a full power down, usually for power center maintenance, we lost at least 2-3 systems. Mostly hard drives, fans and power supplies are the culprit. Sometimes a machine is a running zombie, that is something has gone wrong, like files clobbered or some hardware has failed but its not seen until the machine is rebooted.

Not every program is fully tested and not every system and the intra-system dependencies (system or network wide) are tested fully so coming back to life after the clock has rolled to 1/1/2000 can be an unknown which produces errors that might be false-positives and lead troubleshooters in the wrong direction when trying to solve them.

-- hamster (hamster@mycage.com), December 24, 1999.


Rich,

There is a manufacturing plant in my hometown that is a division of a Fortune 100 company that publicly declared compliance this Summer. The plant processes hazardous chemicals. They didn't complete their remediation, and they have no Y2K contingency plans in place. On December 31, they are doing a complete plant shutdown, which has never been done before. Some of the equipment that is being shut down has been in continuous operation for 10 - 15 years. The person responsible for some of that equipment has told his supervisor that he guarantees that some of that equipment will malfunction on re-start, with or without Y2K problems. The plant is going to go back on-line January 3, one piece at a time, hoping to fix on failure.

-- (RUOK@yesiam.com), December 24, 1999.


I have some embedded chip experience and can see the logic in shutting down over the CDC.

All the real-time-clock chips use two digit years. But they will all roll over from 99 to 00 with no problem. One of the problems will be when the embedded system tries to determine the duration of a time that crosses the CDC. If it starts a timer at 11:00 pm and is supposed to have it expire at 2:00 am it has to know that a date boundary has been crossed. In many cases the RTC handles this. However, at the CDC the "now" time will look earlier than the "before" time. If the system can't deal with this, it will fail.

While shutdown/startup will probably introduce its own risks, the risk of the above can be eliminated by making sure that the system is not running over the CDC. No timers will have to do undefined calculations.

Hope that helps!

-- Gary S. (garys_2k@yahoo.com), December 24, 1999.


Gary S. has part of the answer on embeddeds (I don't know about IT systems). Shutting down over rollover dodges the problem of interval timers that sample the time before and after rollover and calculate a silly interval, with undefined results. This applies only to systems that use dates to calculate intervals, keep real time synchronized with the real world, and present serious problems if they make the wrong decision (to do something they shouldn't or not do something they should).

Another part of the answer has to do with collateral damage. A complex process that gets out of control can do a LOT of damage before anyone can shut it down. So shutting down and bringing back up in a controlled step-by-step way can help identify anything that isn't working correctly before the damage has a chance to occur.

-- Flint (flintc@mindspring.com), December 24, 1999.



Moderation questions? read the FAQ