Exercise

Scheduling a Timer Unit Exercise

Task

Take the service you already wrote and put it on a schedule, then make that schedule survive the machine being switched off. This is the same catch-up behavior anacron gives cron, done per unit.

Host: servera
Prerequisite exercise: A Robust Sync Service
Schedule: every two minutes to test, then daily
Must survive: the machine being switched off

Confirm the service still works

Become root and run the unit from the previous exercise once by hand.
```
sudo -i
systemctl start etc-sync.service
systemctl is-failed etc-sync.service
```
Write a timer for it

A timer takes the name of the service it triggers, with a .timer suffix. That matching name is the whole link between them.
```
vim /etc/systemd/system/etc-sync.timer
```
```
[Unit]
Description=Sync /etc to workstation on a schedule

[Timer]
OnCalendar=*:0/2

[Install]
WantedBy=timers.target
```
```
systemctl daemon-reload
```
This unit does have an [Install] section The timer is the thing that gets enabled, which is why it needs one and the service does not.
Enable the timer, not the service
```
systemctl enable --now etc-sync.timer
systemctl list-timers etc-sync.timer --no-pager
```
NEXT is when it fires next, and ACTIVATES names the service it will start.
Watch it fire

Wait for the next even minute, then look again.
```
systemctl list-timers etc-sync.timer --no-pager
journalctl -u etc-sync.service -n 5 --no-pager
```
LAST and PASSED now show a real run.
Make it survive downtime

First look at what systemd is recording about this timer.
```
ls -l /var/lib/systemd/timers/
```
There is no stamp file, because nothing is tracking the last run yet. Change the schedule to daily and ask for persistence.
```
vim /etc/systemd/system/etc-sync.timer
```
Search with /OnCalendar, replace that line with C, then o to open a line below it:
```
OnCalendar=daily
Persistent=true
```
```
systemctl daemon-reload
systemctl restart etc-sync.timer
ls -l /var/lib/systemd/timers/
```
A stamp file appeared. Count the completed runs, backdate the stamp as though the machine had been off for three days, and restart the timer.
```
journalctl -u etc-sync.service --no-pager | grep -c Finished
touch -d "3 days ago" /var/lib/systemd/timers/stamp-etc-sync.timer
systemctl restart etc-sync.timer
journalctl -u etc-sync.service --no-pager | grep -c Finished
```
The count went up The overdue run happened as soon as the timer came back, without waiting for the next daily slot. That is Persistent=true, and it is what anacron does for /etc/cron.daily, except per unit, without needing root-owned directories and without the battery rule.

Clean up

Remove the timer, the service, the script, and the stamp, on both hosts.

systemctl disable --now etc-sync.timer
rm -f /etc/systemd/system/etc-sync.timer /etc/systemd/system/etc-sync.service
rm -f /usr/local/bin/etc_sync.sh /var/lib/systemd/timers/stamp-etc-sync.timer
systemctl daemon-reload
ssh workstation 'rm -rf /tmp/etc_from_servera'
dnf remove -y rsync

Task

Confirm the service still works

Write a timer for it

Enable the timer, not the service

Watch it fire

Make it survive downtime

Clean up