On 17-Nov-05 Ted Harding wrote:
Just to bring you up to date on my privte saga ... [...] Yet to do: Full test of auto shutdown on "power failure" ( = pull the plug in this case).
But my tentative, and fairly confident, coonclusion is that an itsy bitsy m/c battery will do nicely, at least in emergency.
Power failure test now done. The upshot is that the UPS, on the m/c battery, decides somewhat prematurely that the battery is too low, and simply switches itself off.
Previously, based on timings (when the old battery was in a good state), I had worked out a shutdown programme such that:
1. If power-off detected, wait 2 min (in case it comes back on)
2. The machine with the serial-port connection to the UPS then sends an email to user "shutdown" on the other machine.
3. The other machine has a root cron job running every minute to check whether "shutdown" has mail. If so, this machine immediately executes "shutdown -h now". This takes maximum 2 minutes. So all over for this machine within 5 min of power outage.
4. Meanwhile, the first machine waits a minute (to allow for latency in sending and delivering the mail), and then itself runs "shutdown -h now". Again a minute or two, so all over within 5 minutes.
5. The UPS itself turns off at about 6 minutes, so there's a margin to ensure that the machines are not without power while shutting down.
With the motocycle battery, the UPS decides that the battery is too low at around 2 minutes, and simply turns itself off. The result is that both machines get caught before shutting down.
I suspect that this may be due to the m/c battery having a naturally slightly lower voltage (12.0v) than the UPS expects from a battery in good nick (with the old battery this would be around 12.5v or even a bit higher), so that "dropping" to 12.0v is taken as a sign of debilitation anyway, and the further drop as current is drawn to supply backup power finally does for it.
I'm pondering workrounds, one of which could be to considerably reduce the 2-minute delay at (1) above. I now have enough experience of local conditions to be aware that a power-out is either a few seconds (less than 20 at a rough estimate) or much longer than two minutes. So I might just manage to squeeze that in. The one I can't do much about is the delay plus cron latency in delivering that email to "shutdown", plus the actual shutdown time for the other machine. That could in total go over 2 minutes, which would bring it dangerously close to being powered off before completion.
Hmmm. It's been an interesting experiment, so far, and I'm obliged to everyone who threw in their opinions asnd experiences.
Best wishes to all, Ted.
-------------------------------------------------------------------- E-Mail: (Ted Harding) Ted.Harding@nessie.mcc.ac.uk Fax-to-email: +44 (0)870 094 0861 Date: 17-Nov-05 Time: 23:20:04 ------------------------------ XFMail ------------------------------