More predictable S3 image ID generation
Currently the generated S3 image IDs are completely dependent on insertion order into the s3_images table, which occurs on demand (i.e. these IDs have no global significance).
To facilitate glance image replication (where the glance UUID is maintained across openstack deployments), it would be better if the S3 image ID were made more predictable, modulo collisions.
Ultimately, replication of the s3_images mapping between regions would only be required for the relatively small number of colliding IDs.
Blueprint information
- Status:
- Complete
- Approver:
- Vish Ishaya
- Priority:
- Undefined
- Drafter:
- Eoghan Glynn
- Direction:
- Approved
- Assignee:
- Eoghan Glynn
- Definition:
- Obsolete
- Series goal:
- None
- Implementation:
- Needs Code Review
- Milestone target:
- None
- Started by
- Vish Ishaya
- Completed by
- Vish Ishaya
Related branches
Related bugs
Sprints
Whiteboard
Lots of discussion on the merge prop about whether this approach is correct. Targetted for now until we reach some resolution. --Vish
Looks like based on the discussion that this should be obsoleted. Eoghan? --Vish
Addressed by: https:/
More predictable S3 image ID generation.