On controller failure, some RBD backed Cinder volumes may become unmanageable
Bug #1371328 reported by
Dmitry Borodaenko
This bug affects 2 people
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Mirantis OpenStack |
Fix Committed
|
Medium
|
Dmitry Borodaenko | ||
5.0.x |
Fix Committed
|
Medium
|
Dmitry Borodaenko | ||
5.1.x |
Fix Committed
|
Medium
|
Dmitry Borodaenko | ||
6.0.x |
Fix Committed
|
Medium
|
Dmitry Borodaenko |
Bug Description
According to upstream bug #1028718 and bug #1280367, cinder-volume associates volume metadata with specific host (node where cinder-volume service is running), and will refuse to manage (attach/
Based on comments in bug #1280367, setting host variable in cinder.conf to the same value on all controllers would resolve this limitation as long as multiple backends are not enabled. If that workaround is confirmed, this bug can be downgraded to Medium.
Changed in mos: | |
status: | New → Confirmed |
To post a comment you must log in.
Should this be included into Release notes?
Can the abovementioned workaround be confirmed?