#!/bin/bash

#

# EchoForge Server Stop Script

# This script gracefully shuts down the EchoForge server and cleans up all related processes

# Set strict error handling

set -e

echo "===== Stopping EchoForge Server ====="

echo "$(date): Beginning shutdown process"

# Define log file

LOG_DIR="/home/tdeshane/echoforge/logs"

LOG_FILE="${LOG_DIR}/shutdown_$(date +%Y%m%d_%H%M%S).log"

# Create logs directory if it doesn't exist

mkdir -p "${LOG_DIR}"

# Function to log messages

log() {

local message="$1"

echo "$(date +"%Y-%m-%d %H:%M:%S") - $message" | tee -a "${LOG_FILE}"

}

# Function to check if process exists

process_exists() {

ps -p "$1" > /dev/null 2>&1

return $?

}

# Kill processes by name pattern

kill_processes_by_pattern() {

local pattern="$1"

local signal="$2"

local processes

log "Searching for processes matching pattern: $pattern"

processes=$(pgrep -f "$pattern" 2>/dev/null || true)

if [ -z "$processes" ]; then

log "No processes found matching pattern: $pattern"

return 0

fi

log "Found processes: $processes"

for pid in $processes; do

if process_exists "$pid"; then

log "Sending $signal signal to process $pid ($(ps -p "$pid" -o comm=))"

kill -"$signal" "$pid" 2>/dev/null || log "Failed to send $signal to $pid"

fi

done

}

# Clean up GPU resources

cleanup_gpu() {

log "Checking for GPU processes..."

# Check if nvidia-smi is available

if command -v nvidia-smi >/dev/null 2>&1; then

log "Running nvidia-smi to check GPU usage"

nvidia-smi >> "${LOG_FILE}" 2>&1 || true

# This is a more aggressive approach if needed in the future

# nvidia-cuda-mps-control -d

# echo quit | nvidia-cuda-mps-control

else

log "nvidia-smi not available, skipping GPU cleanup"

fi

}

# Function to ensure port 8765 is released

release_port_8765() {

log "Ensuring port 8765 is released for stable server operation"

# Check for processes using port 8765

if command -v lsof >/dev/null 2>&1; then

PORT_PROCESSES=$(lsof -i :8765 -t 2>/dev/null || true)

if [ -n "$PORT_PROCESSES" ]; then

log "Found processes using port 8765: $PORT_PROCESSES"

for pid in $PORT_PROCESSES; do

log "Terminating process $pid using port 8765"

kill -TERM $pid 2>/dev/null || kill -KILL $pid 2>/dev/null || true

done

else

log "No processes found directly using port 8765"

fi

else

log "lsof command not available, trying alternative methods"

# Alternative using netstat/ss

if command -v ss >/dev/null 2>&1; then

log "Checking socket connections on port 8765"

ss -tanp | grep 8765 >> "${LOG_FILE}" 2>&1 || true

fi

fi

# Wait for socket cleanup

log "Waiting for socket cleanup..."

sleep 3

}

# Main shutdown sequence

main() {

log "Starting EchoForge shutdown sequence"

# 1. Gracefully stop FastAPI/Uvicorn server processes

log "Attempting graceful shutdown of FastAPI/Uvicorn processes"

kill_processes_by_pattern "uvicorn.*app.main" TERM

kill_processes_by_pattern "python.*app.main" TERM

sleep 2

# 2. Stop EchoForge-specific processes

log "Stopping EchoForge-specific processes"

kill_processes_by_pattern "echoforge" TERM

sleep 2

# 3. Force kill any remaining processes if they're still running

log "Checking for remaining processes and force killing if necessary"

kill_processes_by_pattern "uvicorn.*app.main" KILL

kill_processes_by_pattern "python.*app.main" KILL

kill_processes_by_pattern "echoforge" KILL

sleep 1

# 4. Specifically ensure port 8765 is released

release_port_8765

# 5. Clean up GPU resources

cleanup_gpu

# 6. Verify everything is stopped

if pgrep -f "uvicorn.*app.main" >/dev/null 2>&1 || \

pgrep -f "python.*app.main" >/dev/null 2>&1 || \

pgrep -f "echoforge" >/dev/null 2>&1; then

log "WARNING: Some processes may still be running"

else

log "All EchoForge processes successfully terminated"

fi

log "Shutdown sequence completed"

echo "===== EchoForge Server Stopped ====="

}

# Run the main function

main