[5.3] Daemon queue worker: Jobs that exceed timeout will never 'fail'

[maxbrokman]

• Laravel Version: 5.3.6
• PHP Version: 5.6.19|7.0.4
• Database Driver & Version: Mysql (MariaDB 10)
• Queue Driver: Redis, but generally applicable.

Description:

The timeout option when running artisan queue:work --daemon --timeout=n --tries=n successfully kills child processes that run too long but they will never go through the job failure process.

Jobs are marked as failed if they throw an exception during execution and have already been received by a worker too many times. This logic is handled in this method, which is only called from Worker::process() (through handleJobException).

This works well for normal jobs but jobs that exceed the timeout will never actually reach this point, as the parent process kills them while they are still working the logic to 'fail' a job will run.

Steps To Reproduce:

Send jobs to the queue that exceed the timeout and then attempt to work them, for example

public function handle()
{
    echo "Job has been received" . $this->attempts() . " times" . PHP_EOL;
    sleep(10);
}

Working the queue with php artisan queue:work --daemon --timeout=5 --tries=1 I would expect to see this job appear once in output and then fail (ending up the in failed_jobs table etc).

Instead the output is along the lines of

Job has been received 1 times
Job has been received 2 times
Job has been received 3 times
Job has been received 4 times
...

Suggested fix

Move the check for a job that has exceeded its maximum attempts to before the job fires. A job that has repeatedly timed out will then fail the next time the worker receives it.

I'm going to submit a PR for this shortly.

[maxbrokman]

Fixed in #15319 (hopefully :p)