Superpowers Sage:acorn Queues

Queues, Jobs, and Background Tasks

When to Use What

Rule of thumb: Start with Action Scheduler for simple recurring WordPress tasks. Move to Laravel Queue + Job when you need retries, backoff, chaining, batching, or processing that could take more than a few seconds.

Quick Start — Laravel Queue

# 1. Create a job class
bash skills/acorn-queues/scripts/create-job.sh ProcessImage

# 2. Create queue tables (database driver)
lando acorn queue:table
lando acorn queue:failed-table
lando acorn migrate

# 3. Set driver in .env
# QUEUE_CONNECTION=database

# 4. Start a worker
bash skills/acorn-queues/scripts/run-worker.sh
bash skills/acorn-queues/scripts/run-worker.sh emails

Quick Start — Action Scheduler

# Install (if not using WooCommerce)
lando composer require woocommerce/action-scheduler

// Schedule once
as_schedule_single_action(strtotime('+10 minutes'), 'app/sync_content', ['post_id' => 42], 'content-sync');

// Schedule recurring (guard against duplicates)
if (! as_has_scheduled_action('app/cleanup_tokens', [], 'maintenance')) {
    as_schedule_recurring_action(time(), HOUR_IN_SECONDS, 'app/cleanup_tokens', [], 'maintenance');
}

Register callbacks in ThemeServiceProvider::boot():

add_action('app/sync_content', function (int $postId): void {
    $this->app->make(\App\Services\ContentSyncService::class)->syncPost($postId);
});

Job Class Anatomy

class ProcessImage implements ShouldQueue
{
    use Dispatchable, InteractsWithQueue, Queueable, SerializesModels;

    public int $tries = 3;
    public array $backoff = [10, 60, 300]; // exponential: 10s, 1min, 5min

    public function __construct(protected readonly int $attachmentId) {}

    public function handle(): void
    {
        // If job needs WP user context:
        // wp_set_current_user($this->userId);

        $file = get_attached_file($this->attachmentId);
        if (! $file || ! file_exists($file)) {
            return; // idempotent exit — attachment gone
        }
        // ... process
    }

    public function failed(?\Throwable $exception): void
    {
        Log::error('ProcessImage permanently failed', ['attachment_id' => $this->attachmentId]);
    }
}

Dispatching Jobs

// Basic
ProcessImage::dispatch(attachmentId: $id);

// Delayed (5 minutes)
SyncContent::dispatch(sourceId: $id)->delay(now()->addMinutes(5));

// Specific queue
SendEmail::dispatch(campaignId: $id)->onQueue('emails');

// From a WordPress hook
add_action('add_attachment', function (int $id): void {
    if (wp_attachment_is_image($id)) {
        ProcessImage::dispatch(attachmentId: $id);
    }
});

Scripts

# Create a job class (validates PascalCase, checks lando on PATH)
bash skills/acorn-queues/scripts/create-job.sh <JobName>

# Run a queue worker (default queue, 3 tries, 60s backoff)
bash skills/acorn-queues/scripts/run-worker.sh
bash skills/acorn-queues/scripts/run-worker.sh emails

Script source: scripts/create-job.sh · scripts/run-worker.sh

Assets

Boilerplate templates with {{CLASS_NAME}} placeholder. Copy and replace.

• job-simple.php.tpl — Simple job with handle() + failed(), $tries, $backoff, wp_set_current_user stub.
• job-batched.php.tpl — Batched job with use Batchable, $this->batch()->cancelled() check.

References

Deep content loaded on demand — zero tokens until needed.

• action-scheduler.md — Action Scheduler setup, as_schedule_single_action, as_schedule_recurring_action, callbacks, duplicate-schedule prevention, cron-based queue trigger.
• laravel-queue.md — Full config/queue.php, all drivers, database setup, dispatching patterns, job chaining, batching, testing with Bus::fake().
• redis-driver.md — Redis-backed queue, Lando Redis service config (REDIS_HOST=cache), queue isolation, failover to database driver, Supervisor config.
• job-patterns.md — Idempotency patterns, chunking large datasets, retry/backoff config, ShouldBeUnique + uniqueId(), wp_set_current_user in jobs, pass IDs not objects.
• troubleshooting.md — Failed jobs table, queue:work debug flags, common errors (jobs stuck, Redis refused, class not found, memory exhaustion), escalation paths.

Verification

• Dispatch a test job and confirm it appears in the queue: check the jobs table (database driver) or lando redis-cli -h cache LLEN queues:default (Redis driver).
• Run lando acorn queue:work --once and confirm the job processes successfully with expected log output.
• After processing, verify the job is removed from the queue and does not appear in lando acorn queue:failed.

Failure modes

Problem: Jobs dispatched but never processed

• Cause: No worker is running, or QUEUE_CONNECTION in .env does not match a connection in config/queue.php.
• Fix: Start a worker (bash skills/acorn-queues/scripts/run-worker.sh). Verify QUEUE_CONNECTION matches a valid connection key. For the database driver, run lando acorn queue:table && lando acorn migrate.

Problem: Jobs fail and the failed_jobs table does not exist

• Cause: The failed jobs migration was never run.
• Fix: lando acorn queue:failed-table && lando acorn migrate. Then retry with lando acorn queue:retry all.

For all other failure modes see references/troubleshooting.md.

Critical Rules

1. Action Scheduler vs Laravel Queue: Use Action Scheduler for simple recurring WP-native tasks. Use Laravel Queue + Job when you need retries, backoff, chaining, batching, or heavy computation.
2. wp_set_current_user in jobs: If a job calls current_user_can() or any WP function that relies on the current user, call wp_set_current_user($userId) at the top of handle(). Jobs run outside the HTTP lifecycle — WP does not set a current user automatically.
3. Idempotency: Every job must be safe to run more than once. Use guard clauses or upsert patterns. See references/job-patterns.md.
4. Pass IDs, not objects: Pass $postId instead of $post. Re-fetch in handle() to avoid serialization issues and stale data on retry.
5. Set $tries and $backoff: Never let jobs retry infinitely. Use exponential backoff arrays for external API jobs.
6. Failed jobs table must exist: Run lando acorn queue:failed-table && lando acorn migrate before running workers in production.
7. sync driver is for development only. It blocks the HTTP request and runs jobs inline. Set QUEUE_CONNECTION=database or redis in production.
8. Redis host inside Lando is the service name (cache), not 127.0.0.1. See references/redis-driver.md.