dgxmon.sh

#!/usr/bin/env bash
# dgx-monitor.sh — DGX Spark cluster watcher
# Usage: ./dgxmon spark1 spark2 localhost spark3 ...

set -euo pipefail

HOSTS=("$@")
SESSION="dgx-monitor"
REFRESH_MS=2000
USER="${DGX_USER:-$(whoami)}"

if [[ ${#HOSTS[@]} -eq 0 ]]; then
  echo "Usage: $0 host1 host2 ... (use 'localhost' for current machine)"
  exit 1
fi

# Kill existing session if present
tmux kill-session -t "$SESSION" 2>/dev/null || true

# Function to decide command for a host
get_command() {
    local host="$1"

    # Treat these as local
    if [[ "$host" == "localhost" || \
          "$host" == "127.0.0.1" || \
          "$host" == "$(hostname)" || \
          "$host" == "$(hostname -s)" ]]; then
        echo "nv-monitor -i ${REFRESH_MS}"
    else
        echo "ssh -t ${USER}@${host} 'nv-monitor -i ${REFRESH_MS}'"
    fi
}

# Create session with the first host
first_cmd=$(get_command "${HOSTS[0]}")
tmux new-session -d -s "$SESSION" -x "$(tput cols)" -y "$(tput lines)" "$first_cmd"

# Add remaining hosts as split panes
for host in "${HOSTS[@]:1}"; do
    cmd=$(get_command "$host")
    tmux split-window -t "$SESSION" "$cmd"
    tmux select-layout -t "$SESSION" tiled
done

# Select first pane and attach
tmux select-pane -t "$SESSION:0.0"
tmux attach-session -t "$SESSION"