Why does my queue worker still run old code after I deploy?

php artisan queue:work boots the framework once and keeps it in memory. It does not reload code or config until the process restarts. (queue:listen reloads each job but is slower and rarely recommended.)

Fix: restart workers gracefully on every deploy

php artisan queue:restart

This sets a cache timestamp; each worker finishes its current job and then exits. Your process manager (Supervisor, systemd) starts a fresh process with the new code. Requirements:

• A cache driver that all workers can read (not array). With the file driver the web and worker users must share the same path.
• A supervisor entry with autorestart=true.

[program:laravel-worker]
command=php /var/www/app/artisan queue:work --sleep=3 --tries=3 --max-time=3600
autostart=true
autorestart=true
numprocs=2
stopwaitsecs=3600

--max-time=3600 makes workers recycle themselves hourly, which also limits memory leaks.

Failed jobs

Jobs serialize the class name and properties when dispatched. If you rename or change the constructor, old payloads in the queue can fail. Either keep the old signature working for a while, or clear the queue. To retry failed ones after the fix:

php artisan queue:retry all
php artisan queue:failed

Also remember

• php artisan config:cache runs in the deploy script before restarting workers, so they pick up new env values. Never call env() outside config files once config is cached.
• Set retry_after in config/queue.php larger than your longest job timeout or the job runs twice.
• Use $tries, $backoff and a failed() method on the job for sensible retry behaviour.