Cron and queue

The one cron job MagicPost needs, every task it runs, how posts are sent in cron or worker mode, and how to check that it is working.

MagicPost runs all of its background work from the Laravel scheduler. A single cron job that starts the scheduler every minute is enough to publish posts, renew channel tokens, deliver webhooks, read feeds and more. If it does not run, scheduled posts simply stop going out.

Without the cron job nothing is published. Add it as soon as the install finishes.

​

The cron command

MagicPost builds the exact command for your server, with the application path and the PHP binary it found. You can copy it from two places:

• The last screen of the web installer, under One last step: add the cron job.
• Admin → Publishing → Settings, in the Sending card, next to Cron job. Click Copy for the command alone, or Copy the crontab line for the full line.

The crontab line looks like this. Your path and PHP binary will differ, so copy yours from the application.

* * * * * cd /home/USERNAME/public_html && /usr/local/bin/php artisan schedule:run >> /dev/null 2>&1

Under PHP-FPM or LiteSpeed the web server’s PHP is not the command-line PHP. MagicPost uses the command-line php from the same PHP installation when it can find one, and plain php otherwise. If the command it shows does not run, ask your host for the path of the command-line PHP and use that instead.

​

Add the cron job

cPanel

1. Open Cron Jobs

   In cPanel, open Cron Jobs in the Advanced section.

   

   Cron Jobs in the Advanced section of cPanel

2. Choose the interval

   Under Add New Cron Job, open Common Settings and choose Once Per Minute. If your cPanel has no such option, type * in each of the Minute, Hour, Day, Month and Weekday fields.

3. Paste the command

   Paste the command from MagicPost into the Command field and click Add New Cron Job.

Crontab on a VPS

Run crontab -e as the user that owns the application files, paste the full crontab line from the Sending card, and save.

If your host only allows a five-minute interval, the cron job still works, but posts go out up to five minutes late. System health treats the cron job as stopped when the last run is more than five minutes old, so with a five-minute interval it can flag the job between runs.

​

What the scheduler runs

Every task runs inside the schedule:run process, so the scheduler does not need proc_open. Tasks that overlap themselves are guarded with cache locks, which is why System health warns when CACHE_STORE is array or null.

Task	When	What it does
publishing:heartbeat	Every minute	Records the time of the run. System health reads it for the Cron job line.
publishing:dispatch	Every minute	Claims posts that are due and queues them. It also closes off posts that can never go out and posts held back longer than Wait for a channel for up to.
publishing:work:1, publishing:work:2	Every minute, in scheduler mode only	Two lanes that send queued jobs, so one long upload does not stall every other post.
publishing:deliver-webhooks	Every minute	Sends webhook deliveries that are due and clears out old ones.
telegram:poll	Every minute, only while Telegram is set to polling	Fetches waiting Telegram updates for a site that Telegram cannot reach with a webhook. See Set up Telegram.
publishing:recover	Every 5 minutes	Returns posts that were left in the sending state longer than Pick up a stalled send after.
publishing:check-feeds	Every 5 minutes	Reads the RSS feeds that are due and turns new articles into posts.
media:sweep-generations	Every 5 minutes	Picks up AI image and video requests that a crashed run left behind, and releases the credits of requests that never finished.
publishing:recurring	Hourly	Creates the upcoming occurrences of recurring posts.
publishing:refresh-tokens	Hourly	Renews channel access tokens that expire within 24 hours.
publishing:sync-metrics	Hourly, only while Collect analytics is on	Reads follower and post figures from the networks.
publishing:recycle-evergreen	Hourly, at 20 minutes past	Lines up the next repeat of every evergreen post that is due one.
publishing:check-channels	Hourly, at 30 minutes past	Checks channels that have posts due in the next two days, so a revoked token is found before a post fails.
whatsapp:close-abandoned	Hourly	Ends WhatsApp linking sessions that nobody finished.
credits:release-expired-holds	Hourly, at 45 minutes past	Releases credit holds that passed their expiry.
credits:refresh-allowances	Hourly, at 50 minutes past	Refreshes the monthly allowance of yearly and lifetime plans and ends unlimited credits whose plan has lapsed.
publishing:prune-notifications	Daily at 03:15, only while Keep read notifications is not set to For good	Deletes read publishing notifications older than the retention you chose. Unread ones are kept.

​

Cron mode and worker mode

Under Admin → Publishing → Settings, the Sending card has a Posts are sent by choice.

The scheduler

The default, for hosts that only offer cron. Every minute, the two publishing:work lanes process queued jobs for the number of seconds set in Each pass sends for. The value must be between 10 and 55 seconds, and the default is 45, so a pass ends before the next one begins. Lower it if your host stops long cron jobs.

A queue worker

For servers that keep a queue worker running under Supervisor or systemd. The scheduler stops running the publishing:work lanes, and you run this command from the folder MagicPost is installed in:

php artisan queue:work --queue=publishing,default,media

The cron job is still required in this mode, because the scheduler decides which posts are due and queues them.

Both modes use the queue named by QUEUE_CONNECTION in .env, which defaults to database. System health reports the null driver as a problem, because it throws jobs away, and the sync driver as a warning, because each post would then be sent inside the request that queued it. Set QUEUE_CONNECTION=database unless you run another queue backend on purpose.

Hand out at most on the same page limits how many posts are handed to the queue per minute. Posts due at the same moment beyond that limit go out over the following minutes. See Publishing settings for the retry, waiting and upload limits.

​

Check that it works

Open Admin → Publishing → Settings.

• The Sending card shows Last ran with the time of the last scheduler run, or Has not run yet.
• System health shows the Cron job line as passed, with the time of the last run. It fails with “The scheduler has never run” or “The scheduler last ran” when the last run is missing or more than five minutes old.
• The Queue line reports jobs that have waited more than ten minutes, and says whether the cause is a missing worker or a slow scheduler.
• Click Check again after you make a change.

​

Troubleshooting

Has not run yet

The cron job is missing or its command is wrong. Check the path of the application and of the command-line PHP, and make sure the cron user can read the application folder.

The cron job runs but posts stay scheduled

Open System health and read the Queue, Waiting posts and Stalled sends lines. A channel that is paused, needs reconnecting, or belongs to a network that is switched off holds its due posts for up to the Wait for a channel for up to limit.

The Locks line warns about the cache

Set CACHE_STORE=database or CACHE_STORE=file in .env. The array and null stores forget everything between requests, so overlapping scheduler runs cannot be kept apart.

For more, see Scheduler and queue problems.