Comment 7 for bug 1750233

I tried the patch from https://review.openstack.org/565950 as we encountered the issue again recently. The patch does seem to reduce the number of times corruption happens, but it still happens.

I was surprised to see that it looks like a backup is created every time the script is ran. This results in a large number of backups that may contain the same inventory. Checking the inventory for modifications and saving only if needed should reduce the writes that are happening while the backup is taking place in a separate process. Also creating the backup to right before a save, should provide more meaningful backups for administrators.

This is the first time for me digging into the python code on this(and I'm no dev), so I might be missing the reasoning behind the backups being done this way. I put in https://review.openstack.org/#/c/637441/ for these changes. The end result when testing was the inventory and backup was only written to when the conf files were changed. Also any action taken with inventory-manage.py that runs a save will create the backup as well. It also seems to have done away with the corruption issues.