Laravel Forge

On Forge, the autoscaler is just another Daemon. It runs like any other Laravel long-running process.

1. Add the Daemon

In your Forge site: Daemons → Add New Daemon

Field	Value
Command	`php /home/forge/your-app.com/current/artisan queue:autoscale`
User	`forge`
Directory	`/home/forge/your-app.com/current`
Processes	`1`
Stop Wait Seconds	`60`
Stop Signal	`SIGTERM`

Click Create. Forge starts it immediately.

2. Remove any competing Queue Workers

Forge has a separate Queue Workers section. Delete any entry whose queues the autoscaler will manage. Two worker supervisors on one queue will fight:

The autoscaler spawns queue:work --queue=payments
Forge's queue worker UI also spawns queue:work --queue=payments
Both claim the same jobs; unpredictable behaviour follows

You can keep Forge queue workers for queues you list in queue-autoscale.excluded. That's exactly what the excluded config is for — see Queue Topology.

3. Deploys

Forge's default zero-downtime deploy script already restarts Daemons via sudo -S supervisorctl restart, which sends SIGTERM. The autoscaler catches it, gracefully stops spawned workers, and Forge relaunches the Daemon with new code. No extra steps needed.

If you want to restart manually:

# Forge → Daemons → Restart button
# or via SSH:
sudo supervisorctl restart daemon-<daemon-id>
# or inside a deploy hook without sudo:
php artisan queue:autoscale:restart

4. Verify

Forge → Daemons shows the process status. Inspect logs via Forge's log viewer, or SSH:

tail -f /home/forge/.forge/daemon-<id>.log

You should see Autoscale manager started shortly after deploy, then periodic worker-spawn/terminate activity once jobs arrive.

Gotchas

.env changes don't propagate without a Daemon restart. Forge does this automatically after "Update Secrets" in the UI.
PHP version mismatch. Forge servers usually run multiple PHP versions. Make sure the Daemon command uses the same php binary your web layer uses (php8.3 or similar if you pin a specific version).
Long-running manager holds old code until restart. A fresh deploy that changes scaling thresholds won't take effect until the Daemon restarts — which happens automatically on zero-downtime deploy, but not if you skipped that step.