Scrub will report success (exit 0) when the scrub is cancelled

[LordMike]

I've edited the scrub task to report to a healthcheck site when it starts and stops. Just now, I stopped the scrub using btrfs scrub cancel <mount> and noticed the health site reported the task as having successfully run. This must mean an exit code of 0 is returned.

I report like this:

[Service]
# type=simple doesn't seem to fill out $EXIT_STATUS, so we use 'show' instead.
ExecStartPre=-/bin/bash -c "curl -fsS -m 10 --retry 5 -o /dev/null https://hc-ping.com/MY-ID/start"
ExecStopPost=-/bin/bash -c 'curl -fsS -m 10 --retry 5 -o /dev/null https://hc-ping.com/MY-ID/$(systemctl show -p ExecMainStatus btrfs-scrub.service | cut -d= -f2)'

This seems rather unfortunate, so I wanted to report it so that others may know, or it could be fixed. I believe the scrub tasks should actually have continued (run on all filesystems in the loop as it does) but then finally report a non-zero exit if any of the filesystems had an issue.

[LordMike]

I checked out the scrub script, and it surely says:

	if [ "$?" != "0" ]; then
		echo "Scrub cancelled at $MNT"
		exit 1
	fi

I do see Scrub cancelled .. in my journald log, so it should have exited with 1. But if I make my own service like this:

[Service]
Type=simple
ExecStart=bash -c 'sleep 10; exit 1'

It will correctly report 1 for exit code when the service stops:

# systemctl status test
× test.service
     Loaded: loaded (/etc/systemd/system/test.service; static)
     Active: failed (Result: exit-code) since Fri 2024-10-04 14:26:07 CEST; 5min ago
   Duration: 10.008s
   Main PID: 1788283 (code=exited, status=1/FAILURE)
        CPU: 4ms

Oct 04 14:25:57 victoria.home systemd[1]: Started test.service.
Oct 04 14:26:07 victoria.home systemd[1]: test.service: Main process exited, code=exited, status=1/FAILURE
Oct 04 14:26:07 victoria.home systemd[1]: test.service: Failed with result 'exit-code'.

# systemctl show -p ExecMainStatus test.service
ExecMainStatus=1

The btrfs-scrub status:

# systemctl status btrfs-scrub
○ btrfs-scrub.service - Scrub btrfs filesystem, verify block checksums
     Loaded: loaded (/usr/lib/systemd/system/btrfs-scrub.service; static)
    Drop-In: /etc/systemd/system/btrfs-scrub.service.d
             └─override.conf
     Active: inactive (dead) since Fri 2024-10-04 13:52:25 CEST; 39min ago
   Duration: 2min 15.493s
TriggeredBy: ● btrfs-scrub.timer
       Docs: man:fstrim
   Main PID: 1753792 (code=exited, status=0/SUCCESS)
        CPU: 1.712s

Oct 04 13:52:24 victoria.home btrfs-scrub.sh[1753799]: Status:           aborted
Oct 04 13:52:24 victoria.home btrfs-scrub.sh[1753799]: Duration:         0:02:15
Oct 04 13:52:24 victoria.home btrfs-scrub.sh[1753799]: Total to scrub:   4.60GiB
Oct 04 13:52:24 victoria.home btrfs-scrub.sh[1753799]: Rate:             34.88MiB/s
Oct 04 13:52:24 victoria.home btrfs-scrub.sh[1753799]: Error summary:    no errors found
Oct 04 13:52:24 victoria.home btrfs-scrub.sh[1753799]: flock: getting lock took 0.000003 seconds
Oct 04 13:52:24 victoria.home btrfs-scrub.sh[1753799]: flock: executing btrfs
Oct 04 13:52:24 victoria.home btrfs-scrub.sh[1753799]: Scrub cancelled at /mnt/bcached
Oct 04 13:52:25 victoria.home systemd[1]: btrfs-scrub.service: Deactivated successfully.
Oct 04 13:52:25 victoria.home systemd[1]: btrfs-scrub.service: Consumed 1.712s CPU time, 9.5M memory peak, 1.0M memory swap peak.

Note how the main PID reports (code=exited, status=0/SUCCESS), even though the log writes out Scrub cancelled at /mnt/bcached

[LordMike]

Oh!.. The entire script is wrapped in a { .. } code block for logging purposes..

I've altered the exit 0 after all to be:

EXIT_STATUS=${PIPESTATUS[0]}
exit $EXIT_STATUS

And this works. When I cancel my scrub now, I get an exit code 1 in systemd.

[LordMike]

I've prepared a PR that implements this for the four scripts I found that use this piping-logging method.