Bash: tu primer script de comprobación de un servidor LinuxBash: your first health-check script for a Linux server

What you'll get

A Bash script you can drop on any Linux server which, every morning, tells you whether the disk is filling up, a service has died or the load is through the roof. And, along the way, the writing rules that separate a script that helps from one that deletes something one day.

Before you start

• A Linux server (Ubuntu, RHEL, Debian) with SSH access and sudo.
• An editor: nano is fine to start; vim whenever you like.
• A folder of your own to test in: mkdir -p ~/scripts.

1. The five rules

1. Always start the same way: #!/usr/bin/env bash and set -euo pipefail. It stops at the first error (-e), fails on an undefined variable (-u) and does not hide errors inside a pipe (-o pipefail).
2. Quote your variables: "$path", never $path. A space in a file name plus an unquoted rm is how folders get lost.
3. Never rm -rf on a variable without first checking it is not empty and points where you think: [[ -n "$dir" && "$dir" != "/" ]].
4. Read-only first. A script that only reads and warns cannot break anything. Whatever writes (deleting logs, restarting services) is added later, behind an explicit --apply.
5. Run it through shellcheck before executing it on a server. It finds the classic mistakes the eye misses.

2. The script

#!/usr/bin/env bashset -euo pipefail DISK_MAX=85          # % disk usage that triggers a warningLOAD_MAX=4           # 5-minute load averageSERVICES=(sshd cron) # services that must be activeSTATUS=0 warn() { echo "WARN: $*"; STATUS=1; } echo "== $(hostname) · $(date '+%d/%m/%Y %H:%M') ==" # Disk: every real filesystem above the thresholdwhile read -r fs use mount; do  use=${use%\%}  (( use >= DISK_MAX )) && warn "disk $mount at $use% ($fs)"done < <(df -h --output=source,pcent,target -x tmpfs -x devtmpfs | tail -n +2) # Memory and loadread -r _ total used _ < <(free -m | awk '/^Mem:/')echo "memory: $used/$total MB"load=$(cut -d' ' -f2 /proc/loadavg)awk -v l="$load" -v m="$LOAD_MAX" 'BEGIN{exit !(l>=m)}' && warn "5-min load: $load" # Servicesfor s in "${SERVICES[@]}"; do  systemctl is-active --quiet "$s" || warn "service $s is not active"done (( STATUS == 0 )) && echo "OK: no warnings"exit $STATUS

Notes on what it does: df --output avoids counting columns; tail -n +2 skips the header; ${use%\%} strips the %; read ... < <(...) reads from a command without opening a subshell (so STATUS really changes inside the loop). The load comparison goes through awk because Bash does not compare decimals.

3. Test it

›chmod +x ~/scripts/check-server.sh›shellcheck ~/scripts/check-server.sh && echo "shellcheck ok"shellcheck ok›~/scripts/check-server.sh== srv-files-01 · 21/10/2026 08:00 ==WARN: disk /var at 91% (/dev/sda3)memory: 3105/7812 MBWARN: service cron is not active›echo $?1

The exit code (1 with warnings, 0 without) is what lets cron, a monitor or another script know whether it went well without reading the text.

4. Schedule it

›sudo cp ~/scripts/check-server.sh /usr/local/bin/check-server›sudo crontab -e# every day at 07:00, appending to a log that gets rotated0 7 * * * /usr/local/bin/check-server >> /var/log/check-server.log 2>&1›sudo tail -n 20 /var/log/check-server.log

Cron has neither your PATH nor your aliases: use full paths and test the script with env -i bash -c /usr/local/bin/check-server to see how it behaves without your environment. To get an email, put MAILTO=it@contoso.com at the top of the crontab (if the server has mail configured), or better, have it write to the log your monitoring already collects.

Next steps

• Add an --apply that does something useful and safe: rotate a log, restart the dead service with systemctl restart.
• Put the script in the team repository (see the Git guide) so it changes through review, not by hand over SSH.
• In Level 3 you'll see Ansible: when the same script has to run on 40 servers, that is its job.

Is something missing, or different in your environment? Get in touch and I'll fix it.