We have been scratching our heads over the last week trying to debug a large number of course restore failures.
Basically, due to some accidental course deletions, we were restoring about 40 courses from a backup of the system before the courses were deleted.
We were able to extract the courses archives into our running production system. Constantly getting SQL failures similar to
After all sorts of testing and experimentation, we determined that courses would restore with 100% success rate if we disabled cron during the restore. These courses were all of a reasonable size and took from 20 minutes to 5 hours to restore.
Our suggestion is that there is some mechanism for halting cron during course restores. There is a bit of thought involved here as you don't want to:
- kill cron when it is in progress
- lock cron "off" in the instance that course restore process dies
Welcome input and thoughts?