I ran into a problem with the Firefly III package that I think can affect anyone using recurring transactions, so I'm sharing what I found.

What happened

Firefly showed one morning the "cron job hasn't run in 36 hours" warning, and that day's recurring transactions were missing. The app log shows why:

2026-10-05T03:00:01Z taskworker: Starting task ... (app update, package was v3.13.3)
2026-10-05T03:00:46Z apptask: deleteContainer: deleting app containers (app, scheduler)
2026-10-05T03:00:48Z services: setupAddons: setting up addon scheduler with options {"update_feeds":{"schedule":"0 3 * * *", ...}}

The package runs firefly-iii:cron at 0 3 * * *. The auto-update started in the same minute and recreated the scheduler container, so the 03:00 run never happened. There is no ==> cron line for that night. The night before, the cron ran normally at 03:00:22.

Why this matters: Firefly does not catch up

• CreateRecurringTransactions::handleOccurrence() only creates transactions whose date equals the cron's date (if ($date->ne($this->date)) { return null; }). A missed day is not caught up by the next nightly run.
• RecurringCronjob::fire() skips the run if the last run was less than 12 hours ago ($diff <= 43_200). So if you notice the problem in the afternoon and run the cron by hand, the next nightly run at 03:00 skips recurring transactions as well.

I only noticed because of the 36-hour warning. That warning is only visible from about 15:00 until the next night's run, so a missed run is easy to overlook.

Workaround

• Set Cloudron's auto-update time to anything other than 03:00 (System time zone is UTC here; I use 01:00).
• To catch up a missed day (owner account only, because --force requires the owner role):

  sudo -u www-data -E php /app/code/artisan firefly-iii:cron --force --create-recurring --date=YYYY-MM-DD
  

Suggestions for the package

1. Run the cron at a time that doesn't overlap with Cloudron's update windows, or
2. after an update or restart, run the cron once if the last run is more than 24 hours old.

A minor one: the Cloudron health check hits /health every 10 seconds without the X-Remote-User header, and Firefly logs local.ERROR: No user in header "X-Remote-User" each time. That's about 8,600 error lines per day, which buries real errors in the log.

Setup: Firefly III 6.7.7, 4 user accounts (only one in active use), proxy auth via Cloudron.

Thanks!