Exercise

A Robust Sync Service Exercise

Task

Synchronize /etc from servera to workstation when that host is reachable, and fail visibly when it is not. A sync that did not happen is a problem, and systemd will remember it as one.

Hosts: servera and workstation
Sync: /etc to /tmp/etc_from_servera
Runs as: root
Must fail when: workstation is unreachable

Write the sync script

Become root and install rsync, which servera does not have. Remember that rsync must exist on both ends.
```
sudo -i
dnf install -y rsync
```
Write the script with a proper hashbang and its settings as variables, so the host and destination are named once.
```
vim /usr/local/bin/etc_sync.sh
```
```
#!/bin/bash
TARGET_HOST="workstation"
TARGET_DIR="/tmp/etc_from_servera"

rsync -a --delete /etc/ "${TARGET_HOST}:${TARGET_DIR}/"
```
Make it executable and run it once by hand. Never hand systemd a command you have not run yourself.
```
chmod a+x /usr/local/bin/etc_sync.sh
/usr/local/bin/etc_sync.sh
ssh workstation 'ls /tmp/etc_from_servera | head -3'
```
Write a service unit for it

A service unit says how to run something. This one adds a condition that is checked before every run.
```
vim /etc/systemd/system/etc-sync.service
```
```
[Unit]
Description=Sync /etc to workstation

[Service]
Type=oneshot
ExecStartPre=/usr/bin/ping -c1 -W2 workstation
ExecStart=/usr/local/bin/etc_sync.sh
```
Tell systemd to re-read the unit files.
```
systemctl daemon-reload
```
No [Install] section, on purpose Step 4 shows what that costs you, and why this service does not need it.
Start it and read the journal

Start the unit by hand and look at what systemd recorded.
```
systemctl start etc-sync.service
systemctl status etc-sync.service --no-pager
journalctl -u etc-sync.service -n 10 --no-pager
```
Type=oneshot finishes The unit runs to completion and returns to inactive (dead). That is success for a job, not a failure.
Break the check and watch the unit fail

Point the check at an address that does not answer.
```
vim /etc/systemd/system/etc-sync.service
```
Search for the line with /ExecStartPre, press C to replace it from the cursor to the end of the line, and type:
```
ExecStartPre=/usr/bin/ping -c1 -W2 192.0.2.99
```
Reload and start it again.
```
systemctl daemon-reload
systemctl start etc-sync.service
echo "exit status: $?"
systemctl is-failed etc-sync.service
journalctl -u etc-sync.service -n 5 --no-pager
```
The start command exits 1, the unit reports failed, and ExecStart never ran. Now ask systemd what is broken on this machine.
```
systemctl --failed
```
This is what cron cannot do The unit stays in the failed state until it succeeds or you run systemctl reset-failed, so a backup that quietly stopped working is visible in one command. cron mails root once and forgets, which makes a broken job look exactly like a working one.
Restore the check

Put the reachable host back, so the service works for the next exercise.
```
vim /etc/systemd/system/etc-sync.service
```
Search with /ExecCondition, replace the line with C:
```
ExecStartPre=/usr/bin/ping -c1 -W2 workstation
```
```
systemctl daemon-reload
systemctl start etc-sync.service
systemctl is-failed etc-sync.service
systemctl --failed
```
Leave this in place The timer exercise schedules this same unit and cleans up both when it finishes.

Task

Write the sync script

Write a service unit for it

Start it and read the journal

Break the check and watch the unit fail

Restore the check