Quanamo — warmup

quanamo-warmup brings odin back online after a Quanamo deployment. Permissions, config cache, PHP-FPM reload, supervisor processes, Horizon (admin + kanban), kanban-scheduler, scheduler cron, per-app HTTP probe.

Sequence

StepAction
0. Reboot check/var/run/reboot-required → sudo reboot + wait ~30-60s + verify SSH
1. PermissionsPer project: chown -R odin:www-data, chmod -R 775 storage bootstrap/cache, chmod 644 on cached PHP files
2. Config cachePer project: php artisan config:cache && route:cache && view:cache (view:cache tolerated to fail), chmod + chgrp on the bootstrap/cache results
3. PHP-FPMsystemctl reload php8.5-fpm
4. Log cleanupWipe stale laravel*.log + horizon.log per app, truncate supervisor *.log
5. artisan upadmin + feedz + kanban
6. Supervisorsupervisorctl start all. If Horizon needs fresh code: horizon:terminate on admin AND kanban, then start horizon + start kanban-horizon
7. kanban supervisor programsEnsure storage/step-dispatcher/ exists with right ownership. Start kanban-scheduler + kanban-horizon. Do NOT touch active.flag
8. Scheduler cron (admin only)Install /etc/cron.d/quanamo-scheduler — * * * * * odin cd /home/odin/admin.quanamo.com && php artisan schedule:run
9. Verifysupervisorctl status + schedule:list per app + per-app HTTP curl

kanban active.flag — package-managed

Under step-dispatcher v1.12+ the active.flag is package-managed. The package writes the flag when a Step is enqueued, wipes it when the queue drains. Warmup does NOT touch the flag.

If active.flag is absent on an otherwise healthy system, that is the expected idle state — the queue is drained. Enqueue any Step and the flag reappears in the same tick. The old "re-arm during warmup" rule has been retired — see kanban.quanamo.test/deploy/README.md for the lifecycle.

Hard rules

  • Never warmup before deploy health check passes.
  • All project commands as odin, system commands via sudo.
  • Do NOT touch active.flag. Package-managed.
  • Cancel all scheduled tasks after warmup completes.