Backup API allows to use backup taken from another instance as parent for creating incremental backup from another instance

[Denis M.]

Backup APi allows to use backup taken from another instance as parent for creating incremental backup from another instance.

Steps to reproduce:

1. Create an instance. Verify it's active.
 (trove create mysql_first 2 --size 2)
2. Take backup from it. Verify it's completed.
(trove backup-create backup_from_mysql_first <mysql_first>.id)
3. Create another instance. Verify it's active.
 (trove create mysql_second 2 --size 2)
4. Try to create incremental backup.
 (trove backup-create incremental_from_mysql_second <mysql_second>.id --parent-id <backup_from_mysql_first>.id)

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/108101

[Nikhil Manchanda (slicknik)]

Moving bugs that are Medium - Low importance out since they should not block the juno-rc1 release.

[OpenStack Infra (hudson-openstack)]

Change abandoned by Denis M. (<email address hidden>) on branch: master
Review: https://review.openstack.org/108101

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/130522

[Amrith Kumar (amrith)]

Updating importance to "Low" (bugscrub, 2014-10-23)

[OpenStack Infra (hudson-openstack)]

Change abandoned by Nikhil Manchanda (<email address hidden>) on branch: master
Review: https://review.openstack.org/130522
Reason: Abandoning changes that are greater than 1 month old with a negative
review (and no new patchsets) as part of a regular clean up of
reviews. Please feel free to restore your change if this is still work
in progress.

[Doug Shelley (0-doug)]

Wanted to record some commentary on this bug so it doesn't get forgotten...

amrith had this comment during code review:
While it appears to fix the problem you are setting out to fix, it has a side-effect that I believe is problematic.
I create an instance and take a backup. I then launch a new instance based on that backup. I wish to take an incremental backup, based on the backup that I used to launch the instance. It will fail with this implementation.
This is a very real use case; something which I do regularly (not with trove at this time).
Create a database instance. Load a large dataset. Create a backup.
Launch a new instance based on backup. Make some changes to data. Take an incremental backup.
etc ...

I tested this use case with the current code and the restore works (which is a good case and very useful). However, if you do a similar operation but the instance you incremental backup isn't derived from the parent, xtrabackup will fail and the instance will go to error state. Just wondering if there is some way to capture the different cases and validate or invalidate at backup time.