My setup: in my home office I have a rack server, and a desktop plugged into an APC BackUPS Pro 1300. FreeBSD on the server, Win 11 on the desktop. There is also a WiFi and router. The server is a massive power hog, and is only a backup for my primary Co-Lo'ed elsewhares, so I want it to shutdown fairly soon after powerfail. I want the router and WiFi to remain available as long as possible so that my cellphone etc. can use the DSL, assuming of course that it's still up.

I'm now using apcupsd on all systems. The BBU USB cable goes into the desktop as it remains powered longest, UPSCABLE usb, UPSTYPE usb, DEVICE [empty] -- I have it set to power down early (TIMEOUT 4) minutes as this machine takes a significant amount of power which would drain the battery for no real reason. If I were to be at the desktop during that powerfail, this would be sufficient time for me to close up what I was doing. This machine is configured to provide outgoing service (NISIP 0.0.0.0 and NISPORT 3551).

The server is configured to read the desktop for it's BBU info (UPSCABLE ether, UPSTYPE net amd DEVICE [desktopfqdn]. It's set for shutdown very early (TIMEOUT 2) mins. Note that the FreeBSD server will shut down to HALT, but not actually power down by default, making all this useless. To fix this, you must edit /usr/local/etc/apcupsd/apccontrol, find and replace the shutdown -h with shutdown -p.

In the basement where the phone line comes into the house, I have a small APC BBU with only the CenturyLink ActionTec DSL modem plugged into it. No computers, no BBU control wire - it just goes until it runs out of battery. The assumption is that the DSL modem will resume safely when the power returns. This takes very little power and will run for many hours.

In another location, I have another desktop and a Raspberry Pi plugged into another APC Pro 1300. The desktop runs a weather station, but the weather station can handle saving it's own data for about a month, so this desktop is non-critical. It's set to (UPSCABLE ether, UPSTYPE net, DEVICE [rPifqdn]), and to shutdown early, namely TIMEOUT 3. The rPi runs my driveway monitor, thus is considered critical and should run as long as possible. Thus the BBU cable goes into the rPi (UPSCABLE usb, UPSTYPE usb, DEVICE [empty]), which is set to shut down at 5% (BATTERYLEVEL 5) or MINUTES 2. Note that in the rPi, you do not want to change the shutdown command from -h to -p as you want this to resume upon power return, and there's no consequence once we've shutdown and the battery finally reaches exhausted.

I also have a mechanism for sending myself txt messages. A URL with info goes to the rPi, which runs a custom service which then uses Twilio to send to my phone. The service expects JSON encoded data, hence on the Unix boxes, I added the following to each of the scripts in /etc/apcupsd/:
curl -d '{ "file":"'"$MSG"'" }' http://[mylocalserverurl].
On the Windows boxes, in the C:\apcupsd\etc\apccontrol.bat file, I've updated the POPUP= line to:
curl -d "{ \"file\":\"%command%\" }" http://[mylocalserverurl]

upsd
• Plug a serial cable from the UPS port to a serial port on your FreeBSD box. Default is the lowest numbered serial port (probably labelled "Port 1" on the back of the box).
• cd /usr/ports/sysutils/upsd
• make
• make install
• Goto Autostarting and set up an autostart script for upsd. This script need not save the pid in /var/run, the program does this automatically.
• If necessary, edit /usr/local/etc/upsd.conf. The default serial port is /dev/cuaa0 (port 1, the lowest numbered).
• Now you'll get messages on the console that say "Cannot put UPS in smart mode". I guess this means I should have spent more money on my UPS. Mine is an APC BackUPS 500. Oh well...
• make clean
• make deinstall

bkpupsd (Pretty much the same install as upsd)
• Plug in control port.
• cd /usr/ports/sysutils/bkpupsd
• make
• make install
• Goto Autostarting and set up an autostart script for upsd. This script need not save the pid in /var/run, the program does this automatically.
• If desired, edit /usr/local/libexec/bkpups.*. These scripts are executed by the program when events happen.
• After you've done the install, you can use man 8 bkpupsd for more info.

I had to make a special cable to adapt my APC Back-UPS 500 work with this program. A DB-9 male to female (male is UPS side), with all pins just passing through, except:
pin 4 is left open on both sides, and pins 6 and 7 are shorted. The reason is that the UPS is asserting DTR (pin 4) and DSR. DTR is supposed to be driven by the computer end, thus it is being held low by the computer. This in turn, is holding DSR down as well, thus it is left open.

I also had to modify the program -- immediately before the line ioctl(fd, TIOCMBIS, &dtr_bit);, I had to add dtr_bit = -1; this is the first one inside the while(1) loop. Also, the line else if(lstatus & TIOCM_RNG) { gets the TIOCM_RNG replaced with 0x100.
Diff listing:

90a91
> dtr_bit = -1;
112c113
>          else if(lstatus & 0x100 /*TIOCM_RNG*/) {
---
<          else if(lstatus & TIOCM_RNG) {

• The "test" switch causes the same action (as far as the serial port is concerned) as does pulling the power plug.
• Depressing the "test" switch causes the serial status to change, and remain in that state as long as the button is held depressed. When the button is released, the status will return to normal.
• Pulling the power or depressing the test switch with no load on the UPS (nothing plugged in/turned on), the alarm tone will be continuous. With a load, the alarm will beep for about 1 second, about every 4 seconds.
• The UPS asserts DSR and DTR. The UPS also asserts RD. I know, this is bogus (DTR is DCE, DSR and RD are DCE), but that's the behaviour I observed from mine.
• DSR tracks DTR. If your computer asserts DTR (as mine does), this makes it so that no change is observable on DSR either.
• Upon "test" button or powerfail, RD, DSR, and DTR are all forced low. This means that by monitoring DSR you can test for powerfail/test-button.
• If your system were to monitor DTR, instead of driving it, you could monitor DTR to test for powerfail.
• You should also be able to test for powerfail by checking the recieved data for BREAK (eg. framing error).
• Just for the record, my APC 500 held up my computer (4 SCSI HDs spun down, full of RAM, Pentium II 266, internal CDROM idle, nothing else special) for 27 minutes. After that time, the alarm beep became continuous, and after another about 90 seconds, it shut off.
• Due to the amount of time it takes to do a battery charge/discharge cycle, I only did one test for low battery indication. However, the test was conclusive -- I did not observe any status change, meaning it is not possible to detect this status.