launcher/host-scripts/stop-dgx-services.sh

#!/usr/bin/env bash
# stop-dgx-services.sh — Stop non-essential services on a DGX node.
# Frees unified memory for inference. Machine-specific (DGX), not model-
# specific — safe to run before any large-model serve.
#
# Baked into the sm121 binary at compile time (see src/builtin_scripts.rs);
# recipes reference it by alias `stop_dgx_services` in a `before_serve` /
# `after_stop` list. The launcher uploads and runs it via SSH on each node.
#
# To restore services afterwards, use `restore_dgx_services` (or reboot).

set -euo pipefail

SERVICES_STOP=(
    snapd.service
    fwupd.service
    udisks2.service
    upower.service
    ModemManager.service
    bluetooth.service
    avahi-daemon.service
    cups.service
    cups-browsed.service
    colord.service
    smartmontools.service
    rsyslog.service
    rtkit-daemon.service
    dgx-dashboard-admin.service
    dgx-dashboard.service
)

USER_KILL=(
    "pipewire"
    "pipewire-pulse"
    "wireplumber"
    "snapd-desktop-integration"
)

echo "=== [$HOSTNAME] Stopping non-essential services ==="

for svc in "${SERVICES_STOP[@]}"; do
    if systemctl is-active --quiet "$svc" 2>/dev/null; then
        echo "  Stopping $svc ..."
        systemctl stop "$svc" 2>/dev/null || true
        systemctl disable "$svc" 2>/dev/null || true
    fi
done

for proc in "${USER_KILL[@]}"; do
    pids=$(pgrep -x "$proc" 2>/dev/null || true)
    if [ -n "$pids" ]; then
        echo "  Killing $proc (PIDs: $pids) ..."
        kill -TERM $pids 2>/dev/null || true
    fi
done

systemctl stop snapd.service snapd.socket 2>/dev/null || true

echo "  Dropping page cache + dentries ..."
sync
echo 3 > /proc/sys/vm/drop_caches 2>/dev/null || true

echo "=== [$HOSTNAME] Done ==="