Skip to content
← Queue Autoscale for Laravel 4.6.1 View on GitHub

Events

Events

All live in Cbox\LaravelQueueAutoscale\Events. See Event Handling for listener patterns.

Event Fires Payload
ScalingDecisionMade Every cycle $decision
SlaBreachPredicted Every cycle during risk $decision
SlaBreached Once per state transition $connection, $queue, $oldestJobAge, $slaTarget, $pending, $activeWorkers, $slaSignal
SlaRecovered Once per state transition $connection, $queue, $currentJobAge, $slaTarget, $pending, $activeWorkers, $slaSignal
WorkersScaled Per spawn/terminate action $connection, $queue, $from, $to, $action, $reason
WorkerExited Once per worker process the manager finds has ended, including during shutdown (since 4.4) $connection, $queue, $group, $pid, $exitCode, $termSignal, $terminationRequested, $uptimeSeconds, $graceful
ClusterScalingSignalUpdated When the leader publishes a new host scaling signal $clusterId, $leaderId, $currentHosts, $recommendedHosts, $currentCapacity, $requiredWorkers, $action, $reason
AutoscaleManagerStarted When a manager process starts $managerId, $host, $clusterEnabled, $clusterId, $intervalSeconds, $startedAt, $packageVersion
AutoscaleManagerStopped When a manager process stops $managerId, $host, $clusterEnabled, $clusterId, $startedAt, $stoppedAt, $reason, $workerCount, $packageVersion
ClusterLeaderChanged When a manager observes the cluster leader change $clusterId, $previousLeaderId, $currentLeaderId, $observedByManagerId, $changedAt
ClusterManagerPresenceChanged When the leader observes managers join/leave the active set $clusterId, $managerIds, $addedManagerIds, $removedManagerIds, $leaderId, $observedByManagerId, $observedAt
ClusterSummaryPublished When the leader publishes a fresh cluster summary $clusterId, $leaderId, $summary, $publishedAt

SlaBreached / SlaRecovered fields

Property Type Meaning
$connection string Queue connection
$queue string Queue name, or the group name for a group
$oldestJobAge / $currentJobAge int The oldest pending job's age as the driver reported it. Always 0 on Amazon SQS
$slaTarget int sla.target_seconds
$pending int Pending jobs when the transition was observed
$activeWorkers int Workers on the workload when the transition was observed
$slaSignal ?SlaSignal The number the transition was decided on (since 4.5). null only on an event constructed by hand

signalSeconds() returns $slaSignal->seconds, or the job age when there is no signal. breachSeconds() (rounded up), breachPercentage(), marginSeconds() (rounded down) and marginPercentage() are computed from it.

SlaSignal (Cbox\LaravelQueueAutoscale\Scaling\SlaSignal, readonly):

Member Meaning
float $seconds The measured wait
SlaSignalSource $source PickupPercentile ('pickup_percentile'), OldestJobAgeFallback ('oldest_job_age_fallback', too few samples) or OldestJobAge ('oldest_job_age', a strategy that measures no pickups, or the head job's wait on a stalled queue)
?int $percentile The configured sla.percentile when the source is PickupPercentile
?float $secondsSinceLastPickup How long ago the newest pickup in the window was, when the source is PickupPercentile
isBreaching(int $target): bool seconds > 0 && seconds >= $target
isWarning(int $target): bool Over 80% of the target without breaching
isInconclusive(int $pending, bool $driverReportsOldestJobAge = false): bool A fallback reading of 0 with jobs pending, on a driver that cannot report the oldest job's age; holds an open breach
withStallEvidence(int $oldestJobAge, int $pending, int $slaTargetSeconds): SlaSignal The head job's wait when no pickup has been recorded for the target while jobs wait; otherwise the signal unchanged
SlaSignal::driverReportsOldestJobAge(string $driver): bool true for redis and database
describe(): string p95 pickup=42.0s or oldest_age=42s

See How breach is decided.

WorkerExited fields

Property Type Meaning
$connection string Queue connection
$queue string Queue name, or the comma-separated queue list of a group worker
$group ?string Group name for a group worker, else null
$pid ?int Process ID
$exitCode ?int Exit code; 128 + signal when a signal ended the process (137 for SIGKILL)
$termSignal ?int Terminating signal, or null when the process exited on its own
$terminationRequested bool The manager had asked the worker to stop
$uptimeSeconds int Seconds from spawn until the exit was noticed
$graceful bool No signal, and exit code 0 or 12 (queue:work's memory-limit exit)

See Event Handling for how common exits read.

Alerting

AlertRateLimiter

Cache-lock-based cooldown helper. Used internally by BreachNotificationPolicy and recommended for custom alert listeners.

namespace Cbox\LaravelQueueAutoscale\Alerting;

readonly class AlertRateLimiter
{
    public function __construct(public int $cooldownSeconds = 300) {}

    public function allow(string $key): bool;  // true = proceed, false = still in cooldown
}

Resolved from the container with cooldown from queue-autoscale.alerting.cooldown_seconds (env var: QUEUE_AUTOSCALE_ALERT_COOLDOWN).

← Previous Scaling Decision Next → Workers