Scout Version

10.22.1

Scout Driver

Typesense

Laravel Version

12.42.0

PHP Version

8.4.1

Database Driver & Version

No response

SDK Version

No response

Meilisearch CLI Version

No response

Description

I encountered a weird issue recently.

I commented my Typesense credentials in my .env and it of course failed to send scout requests on deployment (no surprise here).

But what I found out is that because of how the MakeSearchable job is configured, when the job receives a timeout, it never fails and is still pending in Horizon.

As a result it just blocked my whole queue and nothing happened with queue messages accumulating.

./artisan horizon:forget --all didn't solve anything because all my messages were not failed but pending.

The solution I found was to clear all pending jobs and to change the default scout job:

<?php

// app/Jobs/MakeSearchableWithFailure.php

namespace App\Jobs;

use Laravel\Scout\Jobs\MakeSearchable;

class MakeSearchableWithFailure extends MakeSearchable
{
    /**
     * The number of times the job may be attempted.
     */
    public int $tries = 1;

    /**
     * The maximum number of unhandled exceptions to allow before failing.
     */
    public int $maxExceptions = 1;

    /**
     * Indicate if the job should be marked as failed on timeout.
     */
    public bool $failOnTimeout = true;

    /**
     * The number of seconds the job can run before timing out.
     */
    public int $timeout = 30;
}

In my AppServiceProvider:

use Laravel\Scout\Scout;

// ...

    /**
     * Bootstrap any application services.
     */
    public function boot(): void
    {
        // ...

        Scout::makeSearchableUsing(\App\Jobs\MakeSearchableWithFailure::class);

        // ...
    }

Maybe I wrongly configured horizon and scout, but I couldn't find anything in the scout documentation regarding this phenomenon.

I don't think a temporary timeout should block the whole queue.

Steps To Reproduce

Remove the driver credentials in env vars, and trigger a few dozens of scout imports.