adt / command-lock

v2.0 2026-11-01 10:39 UTC

Requires

• php: ^7.4|^8.0
• symfony/console: ^4.0|^5.0|^6.0|^7.0|^8.0

MIT

• Apps Dev Team

• symfony
• console
• command
• lock

Published on Packagist.org on 2026-11-01 10:39 UTC

README

namespace App\Commands;

use ADT\CommandLock\CommandLock;
use ADT\CommandLock\Storage\FileSystemStorage;
use Exception;
use Symfony\Component\Console\Input\InputInterface;
use Symfony\Component\Console\Output\OutputInterface;

abstract class Command extends \Symfony\Component\Console\Command\Command
{
	use CommandLock;

	private string $locksDir;

	abstract protected function executeCommand(InputInterface $input, OutputInterface $output): int;

	public function setLocksDir(string $locksDir)
	{
		$this->locksDir = $locksDir;
	}

	/**
	 * @throws Exception
	 */
	protected function execute(InputInterface $input, OutputInterface $output): int
	{
		$this->setStorage(new FileSystemStorage($this->locksDir));

		$this->lock();

		$status = $this->executeCommand($input, $output);

		$this->unlock();

		return $status;
	}
}

FileSystemStorage

Locks are taken with flock() on <locksDir>/<key>.lock. The kernel releases the lock when the process exits (even after kill -9), so there are no stale locks. Lock files are never deleted.

Child processes inherit the lock: a process started from the locked command (exec(), system(), proc_open()) keeps the lock until it exits, even after the command itself has finished.

The locks directory and the lock files are created with permissions 0770 and 0660 regardless of umask, so processes of different users sharing a group can use the same locks. Set the setgid bit on the locks directory (chmod g+s) to make new files inherit its group.

A shared locks directory locks across hosts only if its filesystem forwards flock() to a common lock server. GlusterFS mounted via FUSE does: a lock taken on one host blocks the same command on the other hosts, a released or killed holder frees it for them immediately (tested on Gluster 11.1, replica 3 arbiter 1, also from a docker container with a bind mount). When a host crashes without closing its connection, Gluster releases its locks only after network.ping-timeout (42 s by default). NFS depends on the version and the lock daemon – verify it before relying on it.