this post was submitted on 11 Apr 2024
44 points (97.8% liked)

Linux

48157 readers
638 users here now

From Wikipedia, the free encyclopedia

Linux is a family of open source Unix-like operating systems based on the Linux kernel, an operating system kernel first released on September 17, 1991 by Linus Torvalds. Linux is typically packaged in a Linux distribution (or distro for short).

Distributions include the Linux kernel and supporting system software and libraries, many of which are provided by the GNU Project. Many Linux distributions use the word "Linux" in their name, but the Free Software Foundation uses the name GNU/Linux to emphasize the importance of GNU software, causing some controversy.

Rules

Related Communities

Community icon by Alpár-Etele Méder, licensed under CC BY 3.0

founded 5 years ago
MODERATORS
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
44
[Solved] Trouble setting systemd timer (lemmy.dbzer0.com)
submitted 7 months ago* (last edited 7 months ago) by [email protected] to c/[email protected]
20 comments fedilink hide all child comments
 

I'm trying to move away from cron jobs, not that they don't work, but I want to get on with the times and also learn some things.

I created two user timers (and the associated services), one for backing up my data and the second to upload to B2. I'm using two scripts I had in my cron jobs for a few years and they worked without problems. But with systemd timers both scripts fail with exit code 15 (process terminated) and I have no idea why.

I run Debian 12 Bookworm.

Here's the output for the status of the upload service:

> systemctl --user status rclone-up.service
○ rclone-up.service - Run rclone up for b2
     Loaded: loaded (/home/clmbmb/.config/systemd/user/rclone-up.service; disabled; preset: enabled)
     Active: inactive (dead)
TriggeredBy: ● rclone-up.timer

Apr 11 06:10:39 tesla systemd[1698218]: Starting rclone-up.service - Run rclone up for b2...
Apr 11 06:12:18 tesla systemd[1698218]: rclone-up.service: Main process exited, code=killed, status=15/TERM
Apr 11 06:12:18 tesla systemd[1698218]: rclone-up.service: Failed with result 'signal'.
Apr 11 06:12:18 tesla systemd[1698218]: Stopped rclone-up.service - Run rclone up for b2.
Apr 11 06:12:18 tesla systemd[1698218]: rclone-up.service: Consumed 12.811s CPU time.

Also, here's the log created by rclone while running:

2024/04/11 06:10:42 INFO  : integrity.2376: Copied (new)
2024/04/11 06:10:43 INFO  : hints.2376: Copied (new)
2024/04/11 06:10:43 INFO  : nonce: Copied (replaced existing)
2024/04/11 06:10:47 INFO  : config: Updated modification time in destination
2024/04/11 06:10:55 INFO  : index.2376: Copied (new)
2024/04/11 06:11:40 INFO  :
Transferred:      443.104 MiB / 2.361 GiB, 18%, 16.475 MiB/s, ETA 1m59s
Checks:              1503 / 1503, 100%
Transferred:            4 / 19, 21%
Elapsed time:       1m0.8s
Transferring:
 *                                   data/2/2328: 19% /502.259Mi, 2.904Mi/s, 2m19s
 *                                   data/2/2329: 52% /500.732Mi, 10.758Mi/s, 22s
 *                                   data/2/2330: 14% /501.598Mi, 3.150Mi/s, 2m15s
 *                                   data/2/2331:  0% /500.090Mi, 0/s, -

2024/04/11 06:12:18 INFO  : Signal received: terminated

Where should I look to get some more information about what's going on? Why would the service be terminated like that?

LE:

Setting TimeoutSec=infinity inside the [Service] section of the unit file seems to help. Not 100% if it's a good idea, but I'll experiment with it.

all 21 comments
sorted by: hot top controversial new old
[–] [email protected] 2 points 7 months ago* (last edited 7 months ago) (3 children)

Here is my template

sudo cat > /etc/systemd/user/rsync-backup.service <<EOF
[Unit]
Description=do rsync backups with some conditions
# After=network-online.target

[Service]
Type=oneshot
# require a power connection (optional)
# ExecStartPre=sh -c '[ $(cat /sys/class/power_supply/AC/online) = 1 ]'

# require battery over 40%
# ExecStartPre=sh -c '[ $(cat /sys/class/power_supply/BAT0/capacity) -ge 40 ]'

# require the connected network to NOT be "metered"
# ExecStartPre=sh -c '! $(nmcli -t -f GENERAL.METERED dev show | grep -q 'yes')'

ExecStart=/home/user/.local/bin/rsync-backup
# you might add everything you need
# ExecStart=/path/to/something/else

# delete old logs (disabled for testing)
# ExecStartPost=rm -f /var/log/rsync-backups.log
# log the updates
# ExecStartPost=sh -c 'echo "Last backup: $(date)" > /var/log/rsync-backup.log'
# write errors to log
StandardError=file:/var/log/rsync-backups.log

# GUI message
#ExecStartPost=/usr/bin/notify-send -t 0 -a "Backup" "rsync backup finished." "$(output of some command if you want infos about the backup)"

# run with low priority, when idling
# Nice=15
IOSchedulingClass=idle

# when conditions were not met, try again after 15 minutes
# Restart=on-failure
# RestartSec=900

[Install]
WantedBy=multi-user.target
EOF

Timer file:

sudo cat > /etc/systemd/user/rsync-backup.timer <<EOF
[Unit]
Description=do rsync backups with some conditions

[Timer]
OnCalendar=daily
Persistent=true
EOF

(I think the unit is needed)

That is a slightly modified variant of my automatic rpm-ostree system updates which took an hour or so with the help of ChatGPT and a lot of testing around.

Systemd services are lit.

If you add a "repeat when conditions are not met" you need another timer to start it. Like 2 loops, one big loop to start the process, and one small loop to keep trying until conditions are met. I do that with my system updates to prevent them being done

  • with low battery (or even using an AC requirement)
  • over a metered network
  • when the system is busy
[–] [email protected] 3 points 7 months ago (1 children)

I can't see anywhere that a [Timer] section can be added to a service unit. I tried it and it doesn't work:

Apr 11 13:16:15 computer systemd[2593025]: /home/clmbmb/.config/systemd/user/test.service:10: Unknown section 'Timer'. Ignoring.

The service/script ran as it should, but the Timer section is ignored, as you can see.

[–] [email protected] 1 points 7 months ago

Okay good notice. You should put that in a deparate .timer file then

[–] [email protected] 1 points 7 months ago

Thanks for the suggestion. I'll look into this too.

[–] [email protected] 2 points 7 months ago (1 children)

Is it on the horizon for cron to be deprecated?

[–] [email protected] 7 points 7 months ago (1 children)

I don't think so, but I wanted to try systemd timers.

[–] [email protected] 1 points 7 months ago (1 children)

Sure, not trying to second guess you. What do you think of them so far? Is it worth the effort to learn them?

[–] [email protected] 2 points 7 months ago

Yes, for sure. Timers are more versatile than what cron jobs can do.

[–] [email protected] 3 points 7 months ago

Nothing wrong with learning new tricks, but it's worth mentioning on the side that sometimes a cron job is the right tool.

[–] [email protected] 5 points 7 months ago (1 children)

I think I read something about there being timeouts for slow jobs, which backups definitely are. Really unsure though, it has been a while. Try spawning independent worker programs!

[–] [email protected] 4 points 7 months ago

I just read a bit some internet posts and the documentation. I set TimeoutSec=infinity inside the service and I set the timer to run it now. I'll see if that helps.

[–] [email protected] 8 points 7 months ago (1 children)

Can you share your service file?

[–] [email protected] 4 points 7 months ago (2 children) 
[Unit]
Description=Run rclone up for b2

[Service]
Type=oneshot
ExecStart=/zet/Users/radu/backup/rclone_up.sh

[Install]
WantedBy=rclone-up.timer
[–] [email protected] 6 points 7 months ago (1 children)

You could try setting TimeoutStopSec="infinity" for the service. There may be a default timeout for services and its killing rclone before it can finish because the oneshot type is considered "starting" until the program exits.

[–] [email protected] 7 points 7 months ago (1 children)

TimeoutStopSec applies to the ExecStop command, TimeoutStartSec would be the culprit here. I'm not sure why there would be a default timeout of specifically 1:39 minutes though.

[–] [email protected] 6 points 7 months ago

Yeah you're right, it's time for bed lol

The default start timeout is disabled by default for oneshot.

[–] [email protected] 3 points 7 months ago* (last edited 7 months ago) (1 children)

Does your script fork at some point (and might exit before the rsync job is completed)? ~~Because then you need to use Type=forking instead of simple or oneshot, otherwise systemd will start trying to clean up child processes when the script exits.~~

Edit: Actually considering the time span involved Type=forking will not solve your issue because it will timeout, if this is the problem you need to change your script to not do that.

[–] [email protected] 5 points 7 months ago (1 children)

No, my script doesn't fork and I don't think rclone does that either.

Here's the script (pretty simple):

#!/bin/bash

repos=(fotorepo multirepo persorepo appconfigs)

if pidof -x rclone >/dev/null; then
  echo "Process already running"
  exit 1
fi

for repo in "${repos[@]}"; do
    inidate=$(date +'%Y%m%d_%H%M%S')
    /usr/bin/rclone -v --log-file=/backup/borg/logs/${repo}_b2sync_${inidate}.log sync /backup/borg/${repo} b2:${repo}
    if [[ $? -eq 0 ]]; then
      MSGDATE=$(date +'%d/%m/%y %T')
      mesaj="[${MSGDATE}] Upload for ${repo} was successful."
      curl -H "Title: B2 Upload" -H "Tags: arrow_double_up" -d "${mesaj}" "${URL}"
      #sendmsg "[${MSGDATE}] Upload for <b>${repo}</b> was <b><u>successful</u></b>."
    else
      MSGDATE=$(date +'%d/%m/%y %T')
      mesaj="[${MSGDATE}] Upload for ${repo} has failed. Check the logs."
      curl -H "Title: B2 Upload" -H "Tags: warning" -H "prio:high" -d "${mesaj}" "${URL}"
      #sendmsg "[${MSGDATE}] Upload for <b>${repo}</b> has <b><u>failed</u></b>. Check the logs."
    fi
    enddate=$(date +'%Y%m%d_%H%M%S')
    mv /backup/borg/logs/${repo}_b2sync_${inidate}.log /backup/borg/logs/${repo}_b2sync_${enddate}.log
done
[–] [email protected] 2 points 7 months ago* (last edited 7 months ago) (1 children)

Indeed, that all looks fairly innocuous. Just in case, you are sure that you didn't just accidentially kill or killall rclone or bash?

Perhaps wrapping the script in strace might help debug where the offending signal is coming from.

[–] [email protected] 4 points 7 months ago

Just in case, you are sure that you didn’t just accidentially kill or killall rclone or bash?

No. The process runs at night. Only if my dog started learning Linux and tested something! That makes me wonder...