Limit Bandwidth of Volume Copy
Currently, volume copy operations consumes disk I/O bandwidth heavily and may slow down the other guest instances, during:
* new volume creation from a image
* backup
* volume deletion (when dd if=/dev/zero is chosen to wipe)
etc..
("ionice" option is already implemented in some cases, but it is not always usable, e.g. when instances volume access path doesn't go through I/O scheduler of cinder control node.)
This proposes adding conf to limit bandwidth for volume copy (or up/downloading) to mitigate interference to other instance performance.
For example, the limitation can be implemented by using commands' options to limit bandwidth, or putting commands (such as 'dd') into blkio cgroup for throttling.
Blueprint information
- Status:
- Complete
- Approver:
- John Griffith
- Priority:
- Medium
- Drafter:
- Tomoki Sekiyama
- Direction:
- Approved
- Assignee:
- Tomoki Sekiyama
- Definition:
- Approved
- Series goal:
- Accepted for juno
- Implementation:
- Implemented
- Milestone target:
- 2014.2
- Started by
- Tomoki Sekiyama
- Completed by
- John Griffith
Related branches
Related bugs
Sprints
Whiteboard
This review might be useful (has a token bucket impl u can likely use) - if desired.
https:/
Gerrit topic: https:/
Addressed by: https:/
I/O rate limit for volume copy
Addressed by: https:/
Add blueprint for limit-volume-
Addressed by: https:/
I/O rate limit for volume copy with qemu-img convert
Addressed by: https:/
Add cgroups related commands to rootwrap filters
Addressed by: https:/
Avoid OSError in get_blkdev_
Addressed by: https:/
Disable volume copy bandwidth limit for specified disks
Work Items
Dependency tree
* Blueprints in grey have been implemented.