Blueforce-Linux/scripts/diagnostics/bf-diagnostics

#!/usr/bin/env bash
#
# bf-diagnostics - Deep health check for a Blueforce field terminal.
#
# Detailed version of bf-status: SMART, journal errors, docker ps,
# WireGuard state, unattended-upgrades lock, sockets, inodes, textfiles.
# Never prints secrets, keys, or passwords (handshake keys are redacted).
#
# Install: install -m 0755 bf-diagnostics /usr/local/bin/bf-diagnostics
# Usage: bf-diagnostics [--help]
#
set -euo pipefail

usage() {
    cat <<'EOF'
Usage: bf-diagnostics [--help]

Run deep checks: SMART, journal errors, docker ps, wg show,
unattended-upgrades lock state, sockets, inodes, textfile collectors.

Options:
  --help   Show this help and exit
EOF
}

if [[ "${1:-}" == "--help" || "${1:-}" == "-h" ]]; then
    usage
    exit 0
elif [[ $# -gt 0 ]]; then
    echo "ERROR: unknown argument: $1" >&2; usage >&2; exit 2
fi

set +e  # report, don't abort

sec() { printf '\n### %s\n' "$*"; }

sec "HOST"
hostname; uptime; systemctl get-default 2>/dev/null || echo "target: unknown"

sec "SMART"
if command -v smartctl >/dev/null 2>&1; then
    for dev in /dev/sd? /dev/nvme?n1; do
        [[ -b "${dev}" ]] || continue
        echo "--- ${dev} ---"
        smartctl -H -A "${dev}" 2>/dev/null | grep -Ei 'health|reallocated|pending|uncorrect|wear|temperature|power_on' || echo "no SMART attributes parsed"
    done
else
    echo "smartctl not installed; skipping SMART."
fi

sec "JOURNAL ERRORS (last 24h, top 20)"
if command -v journalctl >/dev/null 2>&1; then
    journalctl --since "24 hours ago" -p err --no-pager -o short 2>/dev/null | tail -20 || echo "no journal access"
else
    echo "journalctl not available."
fi

sec "FAILED SYSTEMD UNITS"
if command -v systemctl >/dev/null 2>&1; then
    systemctl --failed --no-pager 2>/dev/null || echo "no failed units"
else
    echo "systemctl not available."
fi

sec "DOCKER PS"
if command -v docker >/dev/null 2>&1 && docker info >/dev/null 2>&1; then
    docker ps --format 'table {{.Names}}\t{{.Image}}\t{{.Status}}' 2>/dev/null || echo "docker ps failed"
    echo "--- restart counts ---"
    docker ps --format '{{.Names}} restarts={{.RestartCount}}' 2>/dev/null || true
else
    echo "docker daemon not reachable."
fi

sec "WIREGUARD (keys redacted)"
if command -v wg >/dev/null 2>&1; then
    wg show 2>/dev/null | sed -E 's/(private-key|public-key|preshared-key): .*/\1: <redacted>/' || echo "no WireGuard interfaces"
    echo "--- latest handshakes ---"
    wg show all latest-handshakes 2>/dev/null || echo "no handshake data"
else
    echo "wg tool not installed."
fi

sec "UNATTENDED-UPGRADES LOCK (must be DISABLED)"
grep -H . /etc/apt/apt.conf.d/20auto-upgrades 2>/dev/null || echo "20auto-upgrades file missing"
for t in apt-daily.timer apt-daily-upgrade.timer; do
    printf '%s: %s\n' "$t" "$(systemctl is-enabled "$t" 2>/dev/null || echo unknown)"
done

sec "SOCKETS (listening)"
ss -ltnp 2>/dev/null || ss -ltn 2>/dev/null || echo "ss not available"

sec "DISK + INODES"
df -h / /var 2>/dev/null
df -i / /var 2>/dev/null | head -5

sec "TEXTFILE COLLECTORS"
ls -l /var/lib/node_exporter/textfile/ 2>/dev/null || echo "textfile dir missing"
for f in /var/lib/node_exporter/textfile/*.prom; do
    [[ -f "$f" ]] || continue
    echo "--- $f ---"; head -10 "$f"
done

sec "SECURITY UPDATES PENDING"
if command -v apt >/dev/null 2>&1; then
    apt list --upgradable 2>/dev/null | grep -ci security || echo "0 security updates (or apt list unavailable)"
else
    echo "apt not available."
fi

sec "REBOOT REQUIRED"
if [[ -f /var/run/reboot-required ]]; then echo "YES"; cat /var/run/reboot-required.pkgs 2>/dev/null || true
else echo "NO"; fi

echo
echo "Diagnostics complete. For remote triage run: bf-support-bundle"
exit 0