Unable to boot VM from volume snapshot if volume AZ is different from compute AZ
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Mirantis OpenStack |
Invalid
|
Medium
|
MOS Cinder | ||
7.0.x |
Opinion
|
Medium
|
MOS Cinder | ||
8.0.x |
Invalid
|
Medium
|
MOS Cinder |
Bug Description
I'm trying create VM from volume snapshot.
It fails because Volume availability zone differs from Compute availability zone.
We have SIGNLE Ceph cluster, which is default availability zone "nova" in Cinder.
For computes we have multiple availability zones (AZ1, AZ2).
Version: MOS 7.0 (Ceph for all, KVM, Neutron+VxLAN)
Steps to reproduce:
1) Deploy MOS7.0 with Ceph
2) Create host aggregate with availability zone AZ1 (or any name differ from "nova")
3) Create volume from image
4) Create volume snapshot
5) Try to create VM from volume snapshot
Actual result:
Error: Failed to perform requested operation on instance "new01-snap", the instance has an error status: Please try again later [Error: Build of instance 43027300-
Expected result:
VM successfully booted
Full nova-compute.log: https:/
Related discussion: https:/
summary: |
- Unable to VM from volume snapshot if volume AZ is different from compute - AZ + Unable to boot VM from volume snapshot if volume AZ is different from + compute AZ |
description: | updated |
tags: | added: area-cinder area-nova |
Roman, so this was sort of expected behavior in Kilo (7.0) :(
There is a patch for Cinder to ease the pain when Cinder is used with Ceph - https:/ /review. openstack. org/#/c/ 217857/ - it was merged to Liberty and effectively introduced a tiny feature (specifically, a new config option), which allows you to specify a fallback AZ.
Changes introducing new config options are generally not back ported to stable branches. And it won't help on its own - you still have to configure Cinder properly.
I'll let Cinder team comment on this, whether we should put this into a MU doe 7.0 or not.