Running a script right before shut-down of primary

[ErikDB87]

Let me start by apologizing for submitting an "issue" for what is actually a question... But I read the documentation, and tried getting help via the mailing list.

I would like to gracefully shut down my router (Unifi Dream Router (UDR)) after my secondary has shut off (so my primary can follow up on the shut-down of the secondary), and before my primary shuts off (or in any case, before the UPS powers down). Unfortunately, the router doesn't seem to support NUT, but I've modified a bash script I found on the Ubiquiti forum, with which I can shut it down as user nut, running command sudo -H -u nut bash -c 'bash /var/lib/nut/NUT-scripts/shutdownrouter.sh'). (I assume that if any script is to be run, it would be run as user nut.)

erik@MinipcLG2:/var/lib/nut/NUT-scripts$ cat shutdownrouter.sh
#!/bin/bash

# Zie https://community.ui.com/questions/How-to-login-into-UniFi-OS-Console-using-shared-SSH-key-from-a-Mac/c177ccca-d67b-43a3-93fa-5c968ade0b81#answer/f5b9106c-0f3f-43a9-b26a-9bb8486e1ed1
# en https://community.ui.com/questions/Consoles-Monitoring-UPS-via-SNMP-for-Graceful-Shutdown/96979c34-7388-48f1-8053-a351c99c1b32#answer/159ab98f-0086-4215-a494-afa2b48a63bb

# Variables
ssh_timeout=30
ping_timeout=5

# Process for shutting down the UniFi-Controller
shutdown_controller() {
        echo "Shutting down the UniFi-Controller..."

        # Establish SSH-Connection using ssh key to the UniFi-Controller and send shutdown-command
        ssh -o ConnectTimeout=$ssh_timeout -i /var/lib/nut/.ssh/id_ed25519 root@192.168.1.1 "ubnt-systool poweroff"

        # Verify, that the Controller did shut down
        ping -c 20 -W $ping_timeout 192.168.1.1

        if [ $? -eq 0 ]; then
                echo "SUCCESS: The UniFi-Controller has shut down properly."
                exit 0
                else
                echo "ERROR: The UniFi-Controller couldn't be shut down."
                exit 1
        fi
}

# Mainroutine
shutdown_controller

But now I'm trying to figure out how to call it, and I'm not sure... I assume in upsmon.conf there should be a line NOTIFYFLAG FSD SYSLOG+WALL+EXEC. But then I wondered whether FSD was maybe too late a flag?
Then I assume upssched.conf plays a role. This is the current setup:

erik@MinipcLG2:/etc/nut$ sudo cat upsmon.conf
RUN_AS_USER nut
MONITOR apcupskelder@localhost 1 upsmon (password) primary


MINSUPPLIES 1
SHUTDOWNCMD "/sbin/shutdown -h +0"
NOTIFYCMD /usr/sbin/upssched
POLLFREQ 2
POLLFREQALERT 1
HOSTSYNC 15
DEADTIME 15
POWERDOWNFLAG /run/nut/killpower
DEBUG_MIN 6

NOTIFYMSG ONLINE    "upsmon-logger: UPS %s on line power"
NOTIFYMSG ONBATT    "upsmon-logger: UPS %s on battery"
NOTIFYMSG LOWBATT   "upsmon-logger: UPS %s battery is low"
NOTIFYMSG FSD       "upsmon-logger: UPS %s: forced shutdown in progress"
NOTIFYMSG COMMOK    "upsmon-logger: Communications with UPS %s established"
NOTIFYMSG COMMBAD   "upsmon-logger: Communications with UPS %s lost"
NOTIFYMSG SHUTDOWN  "upsmon-logger: Auto logout and shutdown proceeding"
NOTIFYMSG REPLBATT  "upsmon-logger: UPS %s battery needs to be replaced"
NOTIFYMSG NOCOMM    "upsmon-logger: UPS %s is unavailable"
NOTIFYMSG NOPARENT  "upsmon-logger: upsmon parent process died - shutdown impossible"

NOTIFYFLAG ONLINE   SYSLOG+WALL+EXEC
NOTIFYFLAG ONBATT   SYSLOG+WALL+EXEC
NOTIFYFLAG LOWBATT  SYSLOG+WALL
NOTIFYFLAG FSD      SYSLOG+WALL+EXEC
NOTIFYFLAG COMMOK   SYSLOG+WALL+EXEC
NOTIFYFLAG COMMBAD  SYSLOG+WALL+EXEC
NOTIFYFLAG SHUTDOWN SYSLOG+WALL+EXEC
NOTIFYFLAG REPLBATT SYSLOG+WALL
NOTIFYFLAG NOCOMM   SYSLOG+WALL+EXEC
NOTIFYFLAG NOPARENT SYSLOG+WALL

RBWARNTIME 43200

NOCOMMWARNTIME 600

FINALDELAY 30

erik@MinipcLG2:/etc/nut$ sudo cat upssched.conf
CMDSCRIPT /etc/nut/upssched-cmd.sh
PIPEFN /etc/nut/upssched.pipe
LOCKFN /etc/nut/upssched.lock

# Starts a timer when the UPS switches to battery power
AT ONBATT * START-TIMER shutdown_timer 1200

# Cancels the shutdown timer when power is restored
AT ONLINE * CANCEL-TIMER shutdown_timer

# Executes immediate shutdown when battery is low
AT LOWBATT * EXECUTE immediate_shutdown

# Starts a timer on communication failure
AT COMMBAD * START-TIMER commbad_timer 1200

# Cancels the communication failure timer when communication is restored
AT COMMOK * CANCEL-TIMER commbad_timer

# Executes shutdown on persistent communication failure
AT NOCOMM * EXECUTE commbad_shutdown

# Executes powerdown on system shutdown
AT SHUTDOWN * EXECUTE powerdown

Then I assume upssched-cmd.sh plays a role. This is the current setup:

erik@MinipcLG2:/etc/nut$ sudo cat upssched-cmd.sh
#!/bin/sh

case $1 in
    shutdown_timer)
        # Log the event and initiate a controlled shutdown
        logger -t upssched-cmd "UPS running on battery for too long, initiating shutdown"
        /usr/sbin/upsmon -c fsd
        ;;

    immediate_shutdown)
        # Log the critical battery status and initiate immediate shutdown
        logger -t upssched-cmd "UPS on battery critical, forced shutdown"
        /usr/sbin/upsmon -c fsd
        ;;

    commbad_timer)
        # Log persistent communication failures and initiate shutdown
        logger -t upssched-cmd "UPS communication failure persists, initiating shutdown"
        /usr/sbin/upsmon -c fsd
        ;;

    commbad_shutdown)
        # Log communication failure and initiate shutdown
        logger -t upssched-cmd "UPS communication failed, initiating shutdown"
        /usr/sbin/upsmon -c fsd
        ;;

    powerdown)
        # Log the execution of the shutdown
        logger -t upssched-cmd "Executing powerdown command"
        ;;

    *)
        # Log unknown commands
        logger -t upssched-cmd "Unrecognized command: $1"
        ;;
esac

Where should I squeeze in bash /var/lib/nut/NUT-scripts/shutdownrouter.sh?

Or should/could I change SHUTDOWNCMD in upsmon.conf to bash /var/lib/nut/NUT-scripts/shutdownrouter.sh and end that script with /sbin/shutdown -h +0?

I know I can toy around with it, but I prefer asking experts for help, rather than maybe creating a situation which is not ideal.

Thanks for anyone's help!

Kind regards,

Erik

[jimklimov]

I think SHUTDOWNCMD is a proper location to call this specifically when your system goes down due to power situation and not in other cases. It may be prudent to make another small script that calls shutdownrouter.sh and then /sbin/shutdown -h +0, so your experiments with the former do not cause your computer to go down.

[ErikDB87]

Thanks for the tip! I've started up small, with a script calling another script with a simple echo message:

erik@MinipcLG2:/etc/nut$ sudo -H -u nut bash -c 'cat /var/lib/nut/NUT-scripts/nutshutdown.sh'
#!/bin/bash

bash /var/lib/nut/NUT-scripts/testscript.sh
erik@MinipcLG2:/etc/nut$ sudo -H -u nut bash -c 'cat /var/lib/nut/NUT-scripts/testscript.sh'
#!/bin/bash

echo "Piep!"
erik@MinipcLG2:/etc/nut$ sudo -H -u nut bash -c 'bash /var/lib/nut/NUT-scripts/nutshutdown.sh'
Piep!

Is there a simple way I can test whether it works (without pulling the physical plug of the UPS)? I tried upsmon -K, but that didn't work:

erik@MinipcLG2:/etc/nut$ sudo upsmon -K
Network UPS Tools upsmon 2.7.4.11303-11303-ge3ef649ea (development iteration after 2.7.4)
Note: A previous upsmon instance is already running!
Usually it should not be running during OS shutdown,
which is when checking POWERDOWNFLAG makes most sense.
Using power down flag file /run/nut/killpower
   0.000242     [D1] debug level is '6'
   0.000275     Power down flag is not set

(Keep in mind that I'm running a custom build, as you might recall, but running sudo /lib/nut/usbhid-ups -DDDDDD -a apcupskelder -k that also didn't produce Piep!.)

[jimklimov]

upsmon -K only checks if the POWERDOWNFLAG file exists, for nutshutdown script or equivalent late-shutdown integration.

And drivername -k goes to tell a particular device to kill power, and does not care about scripts or OS nuances.

You want upsmon -c fsd (while its service/daemon is running).

[ErikDB87]

You want upsmon -c fsd (while its service/daemon is running).

I tried that, and the primary and secondary shut down... I suppose that happened because I didn't reload something, and the old upsmon.conf configuration was still valid? Because this is what I had changed it to:

erik@MinipcLG2:/etc/nut$ sudo cat upsmon.conf
RUN_AS_USER nut
MONITOR apcupskelder@localhost 1 upsmon (password) primary
MINSUPPLIES 1
#SHUTDOWNCMD "/sbin/shutdown -h +0"
SHUTDOWNCMD "bash /var/lib/nut/NUT-scripts/nutshutdown.sh"
NOTIFYCMD /usr/sbin/upssched
POLLFREQ 2
POLLFREQALERT 1
HOSTSYNC 15
DEADTIME 15
POWERDOWNFLAG /run/nut/killpower
DEBUG_MIN 6
NOTIFYMSG ONLINE    "upsmon-logger: UPS %s on line power"
NOTIFYMSG ONBATT    "upsmon-logger: UPS %s on battery"
NOTIFYMSG LOWBATT   "upsmon-logger: UPS %s battery is low"
NOTIFYMSG FSD       "upsmon-logger: UPS %s: forced shutdown in progress"
NOTIFYMSG COMMOK    "upsmon-logger: Communications with UPS %s established"
NOTIFYMSG COMMBAD   "upsmon-logger: Communications with UPS %s lost"
NOTIFYMSG SHUTDOWN  "upsmon-logger: Auto logout and shutdown proceeding"
NOTIFYMSG REPLBATT  "upsmon-logger: UPS %s battery needs to be replaced"
NOTIFYMSG NOCOMM    "upsmon-logger: UPS %s is unavailable"
NOTIFYMSG NOPARENT  "upsmon-logger: upsmon parent process died - shutdown impossible"
NOTIFYFLAG ONLINE   SYSLOG+WALL+EXEC
NOTIFYFLAG ONBATT   SYSLOG+WALL+EXEC
NOTIFYFLAG LOWBATT  SYSLOG+WALL
NOTIFYFLAG FSD      SYSLOG+WALL+EXEC
NOTIFYFLAG COMMOK   SYSLOG+WALL+EXEC
NOTIFYFLAG COMMBAD  SYSLOG+WALL+EXEC
NOTIFYFLAG SHUTDOWN SYSLOG+WALL+EXEC
NOTIFYFLAG REPLBATT SYSLOG+WALL
NOTIFYFLAG NOCOMM   SYSLOG+WALL+EXEC
NOTIFYFLAG NOPARENT SYSLOG+WALL
RBWARNTIME 43200
NOCOMMWARNTIME 600
FINALDELAY 30

I tried running upsmon -c fsd (obviously after a reboot :)), and this time, only the secondary shut down. That's because of the NOTIFYFLAG, I assume. And that the primary didn't shut down this time, supports my theory that some kind or reload was needed.

But I didn't see the Piep! message. Maybe that's because it was run in the background? And I should check some log?

This didn't show anything:

erik@MinipcLG2:/etc/nut$ journalctl -lu nut-server | grep -i piep
erik@MinipcLG2:/etc/nut$ journalctl | grep -i piep
erik@MinipcLG2:/etc/nut$

Sigh, I didn't chmod +x the script that's actually called... Let's try it again!

[jimklimov]

Yeah, an echo Piep goes to stdout of whoever ran it. May be in journalctl nut-monitor if not suppressed/redirected by anyone along the way.

After config changes, upsmon -c reload or systemctl reload nut-monitor is indeed due.

[ErikDB87]

Yeah, an echo Piep goes to stdout of whoever ran it.

Capturing that stdout is easier to monitor, indeed. So now:

erik@MinipcLG2:/var/lib/nut/NUT-scripts$ sudo cat testscript.sh
#!/bin/bash
timestamp=$(date +"%Y-%m-%d_%H-%M-%S")
echo "Piep! - ($timestamp)" > capture.txt 2>&1

Is there a way to temporary make the secondary not shut down while testing? (Preferably without too much fiddling with the settings 😉 Maybe some line(s) that I can comment away, like NOTIFYMSG SHUTDOWN and/or NOTIFYFLAG SHUTDOWN?)

[jimklimov]

Probably you can comment away its SHUTDOWNCMD, or replace it by a similar log emitter/writer, or stop upsmon there while experimenting on the primary...

[jimklimov]

Maybe also set MINSUPPLIES 0 there - pure monitoring stations can do that. Usually they set the "power value" in MONITOR lines for UPSes that do not feed them to 0 too, so not sure how a mix of non-trivial power values with a zero requirement would fit together, but may be worth trying :)

[ErikDB87]

The secondary uses the Openmediavault plugin, so I just disabled it. If the secondary doesn't respond in x time, the primary proceeds anyway, right?

Alas:

erik@MinipcLG2:/var/lib/nut/NUT-scripts$ sudo upsmon -c fsd
Network UPS Tools upsmon 2.7.4.11303-11303-ge3ef649ea (development iteration after 2.7.4)

Fifteen minutes later:

erik@MinipcLG2:/var/lib/nut/NUT-scripts$ ls - pal
totaal 20
drwxr-xr-x 2 nut nut 4096 jul  4 10:43 ./
drwxr-xr-x 5 nut nut 4096 jun 13 15:26 ../
-rwxrwxr-x 1 nut nut 1155 jul  3 08:33 droomsorteerderuitzetten.sh
-rw-rw-r-- 1 nut nut   57 jul  3 19:39 nutshutdown.sh
-rwxrwxr-x 1 nut nut  125 jul  4 10:43 testscript.sh

I had even put in the absolute path in the script:

erik@MinipcLG2:/var/lib/nut/NUT-scripts$ sudo cat testscript.sh
#!/bin/bash
timestamp=$(date +"%Y-%m-%d_%H-%M-%S")
echo "Piep! - ($timestamp)" > /var/lib/nut/NUT-scripts/capture.txt 2>&1

I think the permissions are okay, no?

Edit: I forgot to chmod +x the script that's actually being run... 🙄 Let's try it again!

[ErikDB87]

Alas, again...

erik@MinipcLG2:/var/lib/nut/NUT-scripts$ sudo upsmon -c fsd
Network UPS Tools upsmon 2.7.4.11303-11303-ge3ef649ea (development iteration after 2.7.4)

Fifteen minutes later:

erik@MinipcLG2:/var/lib/nut/NUT-scripts$ ls -pal
totaal 20
drwxr-xr-x 2 nut nut 4096 jul  4 12:29 ./
drwxr-xr-x 5 nut nut 4096 jun 13 15:26 ../
-rwxrwxr-x 1 nut nut 1155 jul  3 08:33 droomsorteerderuitzetten.sh
-rwxrwxr-x 1 nut nut   57 jul  3 19:39 nutshutdown.sh
-rwxrwxr-x 1 nut nut  125 jul  4 10:43 testscript.sh

For reference, these are the relevant files:

erik@MinipcLG2:/var/lib/nut/NUT-scripts$ sudo cat /etc/nut/upsmon.conf
RUN_AS_USER nut
#MONITOR apcupskelder@localhost 1 admin indUStr*a primary # Deleted 03/01/2025
MONITOR apcupskelder@localhost 1 upsmon indUStr*a primary
MINSUPPLIES 1
#SHUTDOWNCMD "/sbin/shutdown -h +0"
SHUTDOWNCMD "bash /var/lib/nut/NUT-scripts/nutshutdown.sh"
NOTIFYCMD /usr/sbin/upssched
POLLFREQ 2
POLLFREQALERT 1
HOSTSYNC 15
DEADTIME 15
POWERDOWNFLAG /run/nut/killpower
DEBUG_MIN 6
NOTIFYMSG ONLINE    "upsmon-logger: UPS %s on line power"
NOTIFYMSG ONBATT    "upsmon-logger: UPS %s on battery"
NOTIFYMSG LOWBATT   "upsmon-logger: UPS %s battery is low"
NOTIFYMSG FSD       "upsmon-logger: UPS %s: forced shutdown in progress"
NOTIFYMSG COMMOK    "upsmon-logger: Communications with UPS %s established"
NOTIFYMSG COMMBAD   "upsmon-logger: Communications with UPS %s lost"
NOTIFYMSG SHUTDOWN  "upsmon-logger: Auto logout and shutdown proceeding"
NOTIFYMSG REPLBATT  "upsmon-logger: UPS %s battery needs to be replaced"
NOTIFYMSG NOCOMM    "upsmon-logger: UPS %s is unavailable"
NOTIFYMSG NOPARENT  "upsmon-logger: upsmon parent process died - shutdown impossible"
NOTIFYFLAG ONLINE   SYSLOG+WALL+EXEC
NOTIFYFLAG ONBATT   SYSLOG+WALL+EXEC
NOTIFYFLAG LOWBATT  SYSLOG+WALL
NOTIFYFLAG FSD      SYSLOG+WALL+EXEC
NOTIFYFLAG COMMOK   SYSLOG+WALL+EXEC
NOTIFYFLAG COMMBAD  SYSLOG+WALL+EXEC
NOTIFYFLAG SHUTDOWN SYSLOG+WALL+EXEC
NOTIFYFLAG REPLBATT SYSLOG+WALL
NOTIFYFLAG NOCOMM   SYSLOG+WALL+EXEC
NOTIFYFLAG NOPARENT SYSLOG+WALL
RBWARNTIME 43200
NOCOMMWARNTIME 600
FINALDELAY 30

erik@MinipcLG2:/var/lib/nut/NUT-scripts$ sudo cat /var/lib/nut/NUT-scripts/nutshutdown.sh
#!/bin/bash
bash /var/lib/nut/NUT-scripts/testscript.sh

erik@MinipcLG2:/var/lib/nut/NUT-scripts$ sudo cat /var/lib/nut/NUT-scripts/testscript.sh
#!/bin/bash
timestamp=$(date +"%Y-%m-%d_%H-%M-%S")
echo "Piep! - ($timestamp)" > /var/lib/nut/NUT-scripts/capture.txt 2>&1

@jimklimov, any idea what might be going wrong? Maybe bash isn't enough (in upsmon.conf)? Maybe it needs to be something like /sbin/bash or /bin/bash or /usr/bin/bash? I don't know enough about Linux, I'm afraid...

[jimklimov]

Got me wondering about unqualified bash too. In fact, with the script now marked executable, it should suffice that you specify just its file name. Looking at code, it uses system(shutdowncmd) so the standard library should call a system shell and pass it the specified string as a command to execute, and if PATH is set well - the bash would be resolved.

The SHUTDOWNCMD should be executed by the part of upsmon that remains running as root just for this purpose, so permissions to anywhere in the system should not be an issue (unless you started upsmon as nut right away, which is not typical).

Are there any clues in its systemd journal?