What is Magento Cron?

Run crontab -l on the Linux user that owns the codebase (usually www-data or your deploy user, NOT root). You should see three lines invoking bin/magento cron:run --group=default|index|consumers. If they’re missing, run bin/magento cron:install. If they’re present but jobs still aren’t running, check var/log/cron.log and var/log/magento.cron.log for the last lines, you’ll typically see a PHP fatal, a permission denied, or a database connection error. The third place to look is SELECT * FROM cron_schedule ORDER BY scheduled_at DESC LIMIT 20, if everything is "missed", cron isn’t executing at all; if it’s "error", individual jobs are throwing.

Two-step fix. Short term: truncate the old rows. DELETE FROM cron_schedule WHERE scheduled_at < NOW() - INTERVAL 7 DAY, this is safe; the cron history is informational, not state. Long term: tune the history-lifetime configs in Stores → Configuration → Advanced → System → Cron. Defaults are history_success_lifetime = 60 minutes and history_failure_lifetime = 600 minutes, but the cleanup cron itself only runs every 10 minutes, on a high-traffic store with hundreds of jobs/minute the table can still grow large. Drop both lifetimes and run the cleanup job more aggressively, or add a nightly DB cleanup script. The root cause is almost always a misbehaving custom cron job that schedules itself far too frequently.

No, they should be the same Linux user. When cron runs as a different user (commonly root while web runs as www-data), every file the cron writes (cache files, generated/ artifacts, log entries, sitemap.xml) ends up owned by the wrong user. The next web request can’t modify those files and you get permission-denied errors that look unrelated to cron. Standard pattern: install the crontab under the web user with sudo -u www-data crontab -e, or whatever user owns var/ and generated/ in your stack. On Docker setups, the Magento container should run as one user end-to-end.

"System cron" (or "host cron") is the Linux scheduler, the daemon that reads /var/spool/cron/<user> and fires jobs at scheduled times. "Magento cron" is the workload Magento publishes to that scheduler: a registry of jobs (defined in module crontab.xml files) plus the runner (bin/magento cron:run) that the system cron invokes once per minute. Magento doesn’t replace system cron, it depends on it. Without a working host crontab there is no Magento cron. The three lines cron:install writes are how Magento connects its job registry to the host scheduler.

Mostly auto-managed, but not entirely. Adobe Commerce Cloud pre-configures the three cron groups (default, index, consumers) to run on the platform’s worker container at one-minute intervals, you don’t install anything. Where you DO own configuration: the workers stanza in .magento.app.yaml for long-running queue consumers (set the right commands.start and resource limits), the cron history-lifetime tuning in admin, and monitoring, New Relic + cron_schedule dashboards are still your team’s responsibility. Adobe doesn’t alert you when jobs go to "missed" or "error".

Two pieces. First, etc/crontab.xml in your module declares the schedule: <config><group id="default"><job name="myvendor_clean_old_carts" instance="MyVendor\Cart\Cron\CleanOldCarts" method="execute"><schedule>0 3 * * *</schedule></job></group></config>. The schedule string is standard cron syntax (here: 03:00 daily). Second, the PHP class with an execute() method does the work, inject any repository/service via constructor and write your business logic. Once the module is enabled and setup:upgrade has run, the next cron tick will pick up the job and start scheduling rows in cron_schedule. Use the default group for short jobs (< 5 minutes), index for indexer-style work, and consumers for queue-message processing.