10 · Project — Backup Script¶

🎥 Video walkthrough¶

A small end-to-end project combining everything from Level 1: variables, control flow, loops, functions, file tests, redirection, text processing, and error handling.

What you'll build¶

A script backup.sh that:

Takes a source directory as an argument
Archives it into a timestamped .tar.gz file
Logs every step (with timestamps) to a log file, while also printing to the screen
Validates its inputs and fails loudly with clear error messages
Reports the final archive size and exits with a meaningful status code

Project layout¶

backup_project/
    backup.sh
    backups/        (created by the script)
    backup.log       (created by the script)

backup.sh¶

#!/usr/bin/env bash
# backup.sh — archive a directory with logging and error handling
set -euo pipefail

# ---- configuration -------------------------------------------------------
BACKUP_DIR="./backups"
LOG_FILE="./backup.log"

# ---- helpers --------------------------------------------------------------
log() {
    # prints to the screen AND appends to the log file, with a timestamp
    local message="$1"
    local timestamp
    timestamp=$(date '+%Y-%m-%d %H:%M:%S')
    echo "[$timestamp] $message" | tee -a "$LOG_FILE"
}

die() {
    log "ERROR: $1"
    exit "${2:-1}"
}

usage() {
    echo "Usage: $0 <source_directory>" >&2
    exit 1
}

cleanup() {
    log "backup script finished (exit code $?)"
}
trap cleanup EXIT

# ---- validate arguments ----------------------------------------------------
if [[ $# -ne 1 ]]; then
    usage
fi

source_dir="$1"

if [[ ! -d "$source_dir" ]]; then
    die "source directory '$source_dir' does not exist"
fi

# ---- prepare -----------------------------------------------------------------
mkdir -p "$BACKUP_DIR"

base_name=$(basename "$source_dir")
timestamp=$(date +%Y%m%d_%H%M%S)
archive_name="${base_name}_${timestamp}.tar.gz"
archive_path="$BACKUP_DIR/$archive_name"

log "starting backup of '$source_dir'"

# ---- do the backup --------------------------------------------------------
if tar -czf "$archive_path" -C "$(dirname "$source_dir")" "$base_name"; then
    log "archive created: $archive_path"
else
    die "tar failed while archiving '$source_dir'"
fi

# ---- report ------------------------------------------------------------------
if [[ -s "$archive_path" ]]; then
    size=$(du -h "$archive_path" | cut -f1)
    log "backup succeeded — size: $size"
else
    die "archive was created but is empty — something went wrong"
fi

# ---- retention: keep only the 5 most recent backups for this source ------------
log "cleaning up old backups (keeping the 5 most recent for '$base_name')"
# shellcheck disable=SC2012
old_backups=$(ls -1t "${BACKUP_DIR}/${base_name}"_*.tar.gz 2>/dev/null | tail -n +6)

if [[ -n "$old_backups" ]]; then
    while IFS= read -r old_file; do
        rm -f "$old_file"
        log "removed old backup: $old_file"
    done <<< "$old_backups"
else
    log "no old backups to remove"
fi

log "all done."
exit 0

Running it¶

chmod +x backup.sh

./backup.sh ./my_project
# [2026-07-18 10:00:00] starting backup of './my_project'
# [2026-07-18 10:00:00] archive created: ./backups/my_project_20260718_100000.tar.gz
# [2026-07-18 10:00:00] backup succeeded — size: 128K
# [2026-07-18 10:00:00] cleaning up old backups (keeping the 5 most recent for 'my_project')
# [2026-07-18 10:00:00] no old backups to remove
# [2026-07-18 10:00:00] all done.
# [2026-07-18 10:00:00] backup script finished (exit code 0)

./backup.sh ./does_not_exist
# [2026-07-18 10:01:00] ERROR: source directory './does_not_exist' does not exist
# [2026-07-18 10:01:00] backup script finished (exit code 1)

How each Level 1 module shows up here¶

Module	Where it's used
02 Variables	`BACKUP_DIR`, `LOG_FILE`, `archive_name`, quoting throughout
03 Control flow	`if [[ ! -d ... ]]`, `if tar ...; then ... else ... fi`
04 Loops	`while IFS= read -r old_file; do ... done` for cleanup
05 Functions	`log`, `die`, `usage`, `cleanup`
06 Files & directories	`mkdir -p`, `-d`, `-s` tests, `basename`/`dirname`
07 Pipes & redirection	`tee -a`, `<<<`, `2>/dev/null`
08 Text processing	`cut -f1`, `tail -n +6`
09 Exit codes & errors	`set -euo pipefail`, `trap ... EXIT`, `die`

How It Actually Works¶

This script's execution is a chain of fork()/execve() calls stitched together by the parent bash process: each external command (tar, date, find, mkdir) becomes a separate short-lived process, while wait() calls between them (implicit in sequential execution) make sure step N fully finishes — including all its buffered writes reaching disk via the kernel's file cache — before step N+1 starts reading its output.

tar streams file data through its own internal buffer as it walks the directory tree with readdir(2)/stat(2) calls; when combined with a compressor like gzip via a pipe, the two run concurrently exactly as described for pipelines elsewhere in this course — tar never waits for the whole archive to be built before compression starts, it's compressed in a continuous stream.

Timestamped filenames built from $(date +%F) rely on command substitution: bash spawns date in a subshell, captures its stdout through an internal pipe, strips the trailing newline, and splices the result into the string being assigned — all of that happens once, at the moment the assignment line executes, so every reference to that variable later in the script uses the frozen value rather than re-invoking date.

Stretch goals¶

Add a --dry-run flag that logs what would happen without creating an archive or deleting anything.
Compress with zstd instead of gzip if it's available, falling back to gzip otherwise (a command -v zstd check).
Extend log() to also write JSON lines ({"time": ..., "msg": ...}) so the log can be parsed by other tools — you'll build a full log-monitoring script around exactly this idea in Level 2's project.

Completing this project means you're ready for Level 2 · Intermediate.