This refers to scheduled backups, where the number of versions kept is set to one. This customer site has about half a dozen very large courses, where the backup system generates vary large files (two examples from the test I ran this morning are 7.8GB and 9.2GB). These are obviously too large to be successfully zipped, and when the backup processes them, it eventually fails with the error like : "Error - backup-moodle2-course-1681-20130205-1034.mbz does not appear to be a valid backup (missing_moodle_backup_xml_file)"

Then (in the log) it says "complete - next execution (the next scheduled backup time)"

So it fails, presumably because it is too big - but then carries on. This in itself is good, as previously, failures like this were stopping the backup schedule completely.

However, the next time that the scheduled course backup is run, all of the old versions of the backups are replaced (remember, we are only keeping one of each) EXCEPT the failed backups. Moodle does not seem to recognise them as valid backups, despite their filenames, so they are NOT deleted. Over the course of several days, the multiple copies of large failed backups that result from this eventually fill up the server disk, and cause teh site to fail.

I know that this relates to the larger subject of handling very large course backups, but can anything be done with the checking process, so that these large failed backup files do not get left to fill up the server?