modelgrid/ts/systemd.ts

/**
 * ModelGrid Systemd Service Manager
 *
 * Handles installation, removal, and control of the ModelGrid systemd service.
 */

import process from 'node:process';
import { promises as fs } from 'node:fs';
import { execSync } from 'node:child_process';
import { logger } from './logger.ts';
import { theme, symbols } from './colors.ts';
import { PATHS, VERSION } from './constants.ts';

/**
 * Systemd service manager for ModelGrid
 */
export class Systemd {
  private readonly serviceFilePath = '/etc/systemd/system/modelgrid.service';

  private readonly serviceTemplate = `[Unit]
Description=ModelGrid - AI Infrastructure Management
After=network.target docker.service
Requires=docker.service

[Service]
ExecStart=/usr/local/bin/modelgrid service start-daemon
Restart=always
RestartSec=10
User=root
Group=root
Environment=PATH=/usr/bin:/usr/local/bin
WorkingDirectory=/opt/modelgrid

[Install]
WantedBy=multi-user.target
`;

  /**
   * Install the systemd service
   */
  public async install(): Promise<void> {
    try {
      // Check if configuration exists
      await this.checkConfigExists();

      // Write service file
      await fs.writeFile(this.serviceFilePath, this.serviceTemplate);
      logger.logBoxTitle('Service Installation', 50);
      logger.logBoxLine(`Service file created at ${this.serviceFilePath}`);

      // Reload systemd
      execSync('systemctl daemon-reload');
      logger.logBoxLine('Systemd daemon reloaded');

      // Enable service
      execSync('systemctl enable modelgrid.service');
      logger.logBoxLine('Service enabled to start on boot');
      logger.logBoxEnd();
    } catch (error) {
      if (error instanceof Error && error.message === 'Configuration not found') {
        throw error;
      }
      logger.error(`Failed to install systemd service: ${error}`);
      throw error;
    }
  }

  /**
   * Start the systemd service
   */
  public async start(): Promise<void> {
    try {
      await this.checkConfigExists();
      execSync('systemctl start modelgrid.service');
      logger.logBoxTitle('Service Status', 45);
      logger.logBoxLine('ModelGrid service started successfully');
      logger.logBoxEnd();
    } catch (error) {
      if (error instanceof Error && error.message === 'Configuration not found') {
        process.exit(1);
      }
      logger.error(`Failed to start service: ${error}`);
      throw error;
    }
  }

  /**
   * Stop the systemd service
   */
  public stop(): void {
    try {
      execSync('systemctl stop modelgrid.service');
      logger.success('ModelGrid service stopped');
    } catch (error) {
      logger.error(`Failed to stop service: ${error}`);
      throw error;
    }
  }

  /**
   * Get status of the systemd service
   */
  public async getStatus(): Promise<void> {
    try {
      // Display version
      logger.log('');
      logger.log(`${theme.dim('ModelGrid')} ${theme.dim('v' + VERSION)}`);

      // Check if config exists
      try {
        await this.checkConfigExists();
      } catch (_error) {
        return;
      }

      // Display service status
      await this.displayServiceStatus();

      // Display container status
      await this.displayContainerStatus();

      // Display GPU status
      await this.displayGpuStatus();
    } catch (error) {
      logger.error(`Failed to get status: ${error instanceof Error ? error.message : String(error)}`);
    }
  }

  /**
   * Display systemd service status
   */
  private async displayServiceStatus(): Promise<void> {
    try {
      const serviceStatus = execSync('systemctl status modelgrid.service').toString();
      const lines = serviceStatus.split('\n');

      let isActive = false;
      let pid = '';
      let memory = '';

      for (const line of lines) {
        if (line.includes('Active:')) {
          isActive = line.includes('active (running)');
        } else if (line.includes('Main PID:')) {
          const match = line.match(/Main PID:\s+(\d+)/);
          if (match) pid = match[1];
        } else if (line.includes('Memory:')) {
          const match = line.match(/Memory:\s+([\d.]+[A-Z])/);
          if (match) memory = match[1];
        }
      }

      logger.log('');
      if (isActive) {
        logger.log(`${symbols.running} ${theme.success('Service:')} ${theme.statusActive('active (running)')}`);
      } else {
        logger.log(`${symbols.stopped} ${theme.dim('Service:')} ${theme.statusInactive('inactive')}`);
      }

      if (pid || memory) {
        const details = [];
        if (pid) details.push(`PID: ${theme.dim(pid)}`);
        if (memory) details.push(`Memory: ${theme.dim(memory)}`);
        logger.log(`  ${details.join('  ')}`);
      }
      logger.log('');
    } catch (_error) {
      logger.log('');
      logger.log(`${symbols.stopped} ${theme.dim('Service:')} ${theme.statusInactive('not installed')}`);
      logger.log('');
    }
  }

  /**
   * Display container status
   */
  private async displayContainerStatus(): Promise<void> {
    try {
      // Try to get container info from docker
      const output = execSync('docker ps --filter "name=modelgrid" --format "{{.Names}}\\t{{.Status}}"', { encoding: 'utf-8' });
      const lines = output.trim().split('\n').filter(l => l.trim());

      if (lines.length === 0) {
        logger.info('Containers: None running');
        return;
      }

      logger.info(`Containers (${lines.length}):`);

      for (const line of lines) {
        const [name, status] = line.split('\t');
        const isUp = status?.toLowerCase().includes('up');

        logger.log(`  ${isUp ? symbols.running : symbols.stopped} ${theme.highlight(name)} - ${isUp ? theme.success(status) : theme.dim(status)}`);
      }
      logger.log('');
    } catch (_error) {
      // Docker might not be running
    }
  }

  /**
   * Display GPU status
   */
  private async displayGpuStatus(): Promise<void> {
    try {
      // Try nvidia-smi
      const output = execSync('nvidia-smi --query-gpu=name,utilization.gpu,memory.used,memory.total --format=csv,noheader,nounits', { encoding: 'utf-8' });
      const lines = output.trim().split('\n');

      if (lines.length === 0) {
        return;
      }

      logger.info(`GPUs (${lines.length}):`);

      for (const line of lines) {
        const [name, util, memUsed, memTotal] = line.split(',').map(s => s.trim());
        const memPercent = Math.round((parseInt(memUsed) / parseInt(memTotal)) * 100);

        logger.log(`  ${symbols.info} ${theme.gpuNvidia(name)}`);
        logger.log(`    Utilization: ${theme.highlight(util + '%')}  Memory: ${theme.info(memUsed)}/${memTotal} MB (${memPercent}%)`);
      }
      logger.log('');
    } catch (_error) {
      // nvidia-smi might not be available
    }
  }

  /**
   * Disable and uninstall the service
   */
  public async disable(): Promise<void> {
    try {
      // Stop if running
      try {
        execSync('systemctl stop modelgrid.service');
        logger.log('Service stopped');
      } catch (_error) {
        // Might not be running
      }

      // Disable
      try {
        execSync('systemctl disable modelgrid.service');
        logger.log('Service disabled');
      } catch (_error) {
        // Might not be enabled
      }

      // Remove service file
      try {
        await fs.unlink(this.serviceFilePath);
        logger.log('Service file removed');
      } catch (_error) {
        // Might not exist
      }

      // Reload systemd
      execSync('systemctl daemon-reload');
      logger.success('ModelGrid service uninstalled');
    } catch (error) {
      logger.error(`Failed to disable service: ${error}`);
      throw error;
    }
  }

  /**
   * Check if configuration file exists
   */
  private async checkConfigExists(): Promise<void> {
    try {
      await fs.access(PATHS.CONFIG_FILE);
    } catch (_error) {
      logger.log('');
      logger.error('No configuration found');
      logger.log(`  ${theme.dim('Config file:')} ${PATHS.CONFIG_FILE}`);
      logger.log(`  ${theme.dim('Run')} ${theme.command('modelgrid config init')} ${theme.dim('to create one')}`);
      logger.log('');
      throw new Error('Configuration not found');
    }
  }
}