Laravel Queue Autoscale

Intelligent, predictive autoscaling for Laravel queues with SLA/SLO-based optimization.

Laravel Queue Autoscale is a smart queue worker manager that automatically scales your queue workers based on workload, predicted demand, and service level objectives. Unlike traditional reactive solutions, it uses a hybrid predictive algorithm combining queueing theory (Little's Law), trend analysis, and backlog-based scaling to maintain your SLA targets.

Features

🎯 SLA/SLO-Based Scaling - Define max pickup time instead of arbitrary worker counts
📈 Predictive Algorithm - Proactive scaling using trend analysis and forecasting
🔬 Queueing Theory Foundation - Little's Law (L = λW) for steady-state calculations
⚡ SLA Breach Prevention - Aggressive backlog drain when approaching SLA violations
🖥️ Resource-Aware - Respects CPU and memory limits from system metrics
🔄 Metrics-Driven - Uses laravel-queue-metrics for queue discovery and all metrics
🎛️ Extensible - Custom scaling strategies and policies via interfaces
📊 Event Broadcasting - React to scaling decisions, SLA predictions, worker changes
🛡️ Graceful Shutdown - SIGTERM → SIGKILL worker termination
💎 DX First - Clean API following Spatie package conventions

Requirements

PHP 8.3+
Laravel 11.0+
gophpeek/laravel-queue-metrics ^1.0.0
gophpeek/system-metrics ^1.2

Installation

Install via Composer:

composer require gophpeek/laravel-queue-autoscale

Publish the configuration file:

php artisan vendor:publish --tag=queue-autoscale-config

Setup Metrics Package

The autoscaler requires laravel-queue-metrics for queue discovery and metrics collection:

# Install metrics package (if not auto-installed via dependency)
composer require gophpeek/laravel-queue-metrics

# Publish metrics configuration
php artisan vendor:publish --tag=queue-metrics-config

Configure storage backend in .env:

# Option A: Redis (recommended - fast, in-memory)
QUEUE_METRICS_STORAGE=redis
QUEUE_METRICS_CONNECTION=default

# Option B: Database (persistent storage)
QUEUE_METRICS_STORAGE=database

If using database storage, publish and run migrations:

php artisan vendor:publish --tag=laravel-queue-metrics-migrations
php artisan migrate

📚 See Metrics Package Documentation for advanced configuration.

Quick Start

1. Configure SLA Targets

Edit config/queue-autoscale.php:

return [
    'enabled' => true,

    'sla_defaults' => [
        'max_pickup_time_seconds' => 30,  // Jobs picked up within 30s
        'min_workers' => 1,
        'max_workers' => 10,
        'scale_cooldown_seconds' => 60,
    ],

    // Per-queue overrides
    'queues' => [
        'emails' => [
            'max_pickup_time_seconds' => 60,  // Less strict SLA
            'max_workers' => 5,
        ],
        'critical' => [
            'max_pickup_time_seconds' => 5,   // Strict SLA
            'max_workers' => 20,
        ],
    ],
];

2. Run the Autoscaler

php artisan queue:autoscale

The autoscaler will:

Receive all queues and metrics from laravel-queue-metrics
Apply scaling algorithms to meet SLA targets
Scale workers up/down based on calculations
Respect CPU/memory limits from system-metrics
Log all scaling decisions

3. Monitor with Events

use PHPeek\LaravelQueueAutoscale\Events\WorkersScaled;
use PHPeek\LaravelQueueAutoscale\Events\SlaBreachPredicted;

Event::listen(WorkersScaled::class, function (WorkersScaled $event) {
    Log::info("Scaled {$event->queue}: {$event->from} → {$event->to} workers");
    Log::info("Reason: {$event->reason}");
});

Event::listen(SlaBreachPredicted::class, function (SlaBreachPredicted $event) {
    // Alert when SLA breach is predicted
    Notification::route('slack', env('SLACK_ALERT_WEBHOOK'))
        ->notify(new SlaBreachAlert($event->decision));
});

How It Works

Hybrid Predictive Algorithm

The autoscaler uses three complementary approaches and takes the maximum (most conservative):

1. Rate-Based (Little's Law)

Workers = Arrival Rate × Avg Processing Time

Calculates steady-state workers needed for current load.

2. Trend-Based (Predictive)

Workers = Predicted Rate × Avg Processing Time

Uses trend analysis to predict future arrival rates and scale proactively.

3. Backlog-Based (SLA Protection)

Workers = Backlog / (Time Until SLA Breach / Avg Job Time)

Aggressively scales when approaching SLA violations.

Resource Constraints

All calculations are bounded by:

System capacity - CPU/memory limits from system-metrics
Config bounds - min/max workers from configuration
Cooldown periods - Prevents scaling thrash

See ARCHITECTURE.md for detailed algorithm explanation.

Configuration Reference

SLA Defaults

'sla_defaults' => [
    // Maximum time a job should wait before being picked up (seconds)
    'max_pickup_time_seconds' => 30,

    // Minimum workers to maintain (even if queue is empty)
    'min_workers' => 1,

    // Maximum workers allowed for this queue
    'max_workers' => 10,

    // Cooldown period between scaling operations (seconds)
    'scale_cooldown_seconds' => 60,
],

Per-Queue Overrides

'queues' => [
    'queue-name' => [
        'max_pickup_time_seconds' => 60,
        'min_workers' => 2,
        'max_workers' => 20,
        'scale_cooldown_seconds' => 30,
    ],
],

Prediction Settings

'prediction' => [
    // How far ahead to forecast (seconds)
    'forecast_horizon_seconds' => 60,

    // When to trigger backlog drain (0-1, e.g., 0.8 = 80% of SLA time)
    'breach_threshold' => 0.8,
],

Resource Limits

'resource_limits' => [
    // Maximum CPU usage percentage
    'max_cpu_percent' => 90,

    // Number of CPU cores to reserve
    'reserve_cpu_cores' => 0.5,

    // Maximum memory usage percentage
    'max_memory_percent' => 85,

    // Estimated memory per worker (MB)
    'worker_memory_mb_estimate' => 128,
],

Worker Settings

'worker' => [
    // Worker process arguments
    'tries' => 3,
    'timeout_seconds' => 3600,
    'sleep_seconds' => 3,

    // Graceful shutdown timeout before SIGKILL
    'shutdown_timeout_seconds' => 30,
],

Strategy Configuration

'strategy' => [
    // Scaling strategy class (must implement ScalingStrategyContract)
    'class' => \PHPeek\LaravelQueueAutoscale\Scaling\Strategies\PredictiveStrategy::class,
],

Custom Scaling Strategies

Implement your own scaling logic:

use PHPeek\LaravelQueueAutoscale\Contracts\ScalingStrategyContract;
use PHPeek\LaravelQueueAutoscale\Configuration\QueueConfiguration;

class CustomStrategy implements ScalingStrategyContract
{
    public function calculateTargetWorkers(object $metrics, QueueConfiguration $config): int
    {
        // Your custom logic here
        return (int) ceil($metrics->processingRate * 2);
    }

    public function getLastReason(): string
    {
        return 'Custom strategy: doubled the processing rate';
    }

    public function getLastPrediction(): ?float
    {
        return null; // Optional: estimated pickup time
    }
}

Register in config:

'strategy' => [
    'class' => \App\Scaling\CustomStrategy::class,
],

Scaling Policies

Add before/after hooks to scaling operations:

use PHPeek\LaravelQueueAutoscale\Contracts\ScalingPolicy;
use PHPeek\LaravelQueueAutoscale\Scaling\ScalingDecision;

class NotifySlackPolicy implements ScalingPolicy
{
    public function beforeScaling(ScalingDecision $decision): void
    {
        if ($decision->shouldScaleUp()) {
            Slack::notify("About to scale up {$decision->queue}");
        }
    }

    public function afterScaling(ScalingDecision $decision, bool $success): void
    {
        if (!$success) {
            Slack::notify("Failed to scale {$decision->queue}");
        }
    }
}

Register in config:

'policies' => [
    \App\Policies\NotifySlackPolicy::class,
],

Events

Subscribe to scaling events:

ScalingDecisionMade

use PHPeek\LaravelQueueAutoscale\Events\ScalingDecisionMade;

Event::listen(ScalingDecisionMade::class, function (ScalingDecisionMade $event) {
    $decision = $event->decision;

    Log::info('Scaling decision', [
        'queue' => $decision->queue,
        'current' => $decision->currentWorkers,
        'target' => $decision->targetWorkers,
        'reason' => $decision->reason,
    ]);
});

WorkersScaled

use PHPeek\LaravelQueueAutoscale\Events\WorkersScaled;

Event::listen(WorkersScaled::class, function (WorkersScaled $event) {
    Metrics::gauge('queue.workers', $event->to, [
        'queue' => $event->queue,
        'action' => $event->action, // 'scaled_up' or 'scaled_down'
    ]);
});

SlaBreachPredicted

use PHPeek\LaravelQueueAutoscale\Events\SlaBreachPredicted;

Event::listen(SlaBreachPredicted::class, function (SlaBreachPredicted $event) {
    $decision = $event->decision;

    // Alert when pickup time is predicted to exceed SLA
    if ($decision->isSlaBreachRisk()) {
        PagerDuty::alert("SLA breach predicted for {$decision->queue}");
    }
});

Advanced Usage

Running as Daemon

Use Supervisor to keep the autoscaler running:

[program:queue-autoscale]
command=php /path/to/artisan queue:autoscale
directory=/path/to/project
user=www-data
autostart=true
autorestart=true
redirect_stderr=true
stdout_logfile=/path/to/logs/autoscale.log

Custom Evaluation Interval

php artisan queue:autoscale --interval=10

Default is 5 seconds between evaluations.

Debugging

Enable detailed logging:

'log_channel' => 'stack', // Use your preferred channel

View scaling decisions:

tail -f storage/logs/laravel.log | grep autoscale

Metrics Integration

This package does NOT discover queues or collect metrics itself. All queue discovery and metrics collection is delegated to laravel-queue-metrics:

use PHPeek\LaravelQueueMetrics\QueueMetrics;

// The ONLY source of queue data for autoscaling
$allQueues = QueueMetrics::getAllQueuesWithMetrics();

foreach ($allQueues as $queue) {
    echo "Queue: {$queue->connection}/{$queue->queue}\n";
    echo "Processing Rate: {$queue->processingRate} jobs/sec\n";
    echo "Backlog: {$queue->depth->pending} jobs\n";
    echo "Oldest Job: {$queue->depth->oldestJobAgeSeconds}s\n";
    echo "Trend: {$queue->trend->direction}\n";
}

Package Responsibilities:

laravel-queue-metrics (dependency)

✅ Scans all configured queue connections
✅ Discovers active queues
✅ Collects queue depth and age metrics
✅ Calculates processing rates
✅ Analyzes trends and forecasts