BlockDeviceDriver uses iscsi protocol to mount volumes
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Mirantis OpenStack |
Fix Released
|
High
|
Nikita Konovalov | ||
6.1.x |
Won't Fix
|
High
|
Nikita Konovalov |
Bug Description
If Cinder is configured to use the BlockDeviceDriver and the block device is located locally to the host with VM it should be attached directly without iscsi target/initiator used.
However it still uses iscsi on the localhost. It results in high CPU load an low disk performance.
Steps to reproduce:
* Change the backing dirver in cinder.conf to BlockDeviceDriver.
* Add local disks to the list of the device in cinder.conf.
See this doc for config parameters' names. https:/
* Restart cinder-volume service where it is set up
* Launch a vm and attach as many volumes to it on the same host as possible.
* Run a disk performance test on host and on VM. The difference will be significant.
Changed in mos: | |
assignee: | nobody → MOS Cinder (mos-cinder) |
Changed in mos: | |
importance: | Undecided → Medium |
Changed in mos: | |
milestone: | none → 7.0 |
Changed in mos: | |
assignee: | MOS Cinder (mos-cinder) → Nikita Konovalov (nkonovalov) |
description: | updated |
Changed in mos: | |
status: | New → Confirmed |
importance: | Medium → High |
Changed in mos: | |
status: | Confirmed → In Progress |
tags: | added: sahara |
Changed in mos: | |
status: | In Progress → Fix Committed |
Changed in mos: | |
status: | Fix Committed → Fix Released |
Nikita, could you please describe steps to reproduce and version of ISO where we have the bug? It will help to properly track the fix during the release and verify the fix when the issue will be fixed.