Cron stops working after a Restore of a Scheduled Backup

Problem

If a scheduled backup is used as the source of a restore, Drupal's cron will always fail on the restored site, and so no further backups (or other cron initiated actions) will occur.

Cause

This happens because scheduled backups are invoked by Drupal's cron.php which sets the Drupal variable cron_semaphore to indicate it is running, and clears it once completed. This is used to prevent multiple cron jobs from executing at the same time. Therefore when Backup and Migrate dumps the database it will include the cron_semaphore variable.

When the database is restored, a row in Drupal's variable table will be created for the cron_semaphore variable. Since the restore is not being done when Drupal's cron is running, no one will delete the cron_semaphore row and so cron will no longer execute.

This situation can be checked by going to Administer->Reports->Status report and manually invoking cron. It will report "Attempting to re-run cron while it is already running." in the status. The Administer->Reports->Recent log entries will also contain the same error entry.

Suggested Solution

It seems that Backup and Migrate's restore should delete the cron_semaphore variable after the restore if it is present (restoring from a manual backup will not have it set). Also, if the option to put the site offline was selected, then the restored site will be offline. It might be helpful to warn the user of this after the restore so they can go and manually put it back online.

If mysql or phpMyAdmin is used to do the restore, the same problem will occur. So it would probably be worth documenting in the help page.

None of this would be necessary if Drupal's core had a better method of determining if cron was running. For example, it could record the start time and a pid of the cron process. If the semaphore and start time indicated cron had been running a long time, then the pid could be used to see if it was really running, and clear the semaphore if not. That way cron would reset itself if the job crashed or a database restore had been performed.

Work Around

If a database is restored and cron will no longer execute, simply delete the cron_semaphor row in the variable table and clear Drupal's cache.

Using phpMyAdmin access your database and browse the "variable" table. Find the row with the "name" of "cron_semaphor" and delete it. Then go to Administer->Site configuration->Performance and click "Clear cached data". To confirm this has fixed the problem go to Administer->Reports->Status report and manually invoking cron. It should report that it completed successfully.

The cron_semaphor variable can also be deleted using the devel module if it is installed.