block migration fails when disk_allocation_ratio>1.0

Bug #1659539 reported by Timofey Durakov
8
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Mirantis OpenStack
Fix Released
High
Timofey Durakov

Bug Description

When disk_allocation_ratio config option > 1.0 and no shared storage is used, migration fails on checks with:
ERROR (BadRequest): Migration pre-check error: Unable to migrate 9f5360f3-282f-4a50-983c-7b0025d9509f: Disk of instance is too large(available on destination host:-727996956672 < need:65929216) (HTTP 400) (Request-ID: req-ef12922d-5032-4b59-8963-4ce124102662)

The root cause for that is that nova compares disk space, required by instance, with disk_available_least field, which is possible to be negative, due to overcommit.

Changed in mos:
importance: Undecided → High
status: New → In Progress
Revision history for this message
Timofey Durakov (tdurakov) wrote :
Revision history for this message
Sergii Rizvan (srizvan) wrote :

Note for QA:

In order to verify the fix we need to create situation when disk_available_least value on some hypervisor will be negative due to overcommit and try to migrate some instance to that hypervisor.
Steps for verification with the CLI commands described by the following link: http://paste.openstack.org/show/598428/

Sergii Rizvan (srizvan)
tags: added: customer-found
Changed in mos:
milestone: 8.0-updates → 8.0-mu-4
Revision history for this message
Fuel Devops McRobotson (fuel-devops-robot) wrote : Fix merged to openstack/nova (openstack-ci/fuel-8.0/liberty)

Reviewed: https://review.fuel-infra.org/30121
Submitter: Pkgs Jenkins <email address hidden>
Branch: openstack-ci/fuel-8.0/liberty

Commit: 0a303ddb9a70c79f8998b6c5a7d5fb3291d5136d
Author: Timofey Durakov <email address hidden>
Date: Thu Jan 26 12:26:23 2017

fix for disk-allocation ratio during block-migration

for the block-migration cases nova does extra check on
compute node to honor disk_over_commit api parameter,
which lead compute to do this check without taking into
account disk_allocation_ratio config option. This leads
nova to report, that there are no disk space on destination
node

This patch introduce new check on conductor side for the case,
when operator passed target host over api, and make check on
compute side optional, depending on disk_over_commit parameter

Closes-bug: #1659539

Change-Id: I430ca07586343dc71aef7fe9e440f21fc8923d4a

Changed in mos:
status: In Progress → Fix Committed
Revision history for this message
Ekaterina Shutova (eshutova) wrote :

Verified on MOS 8.0 mu4 updates.
Block migration of instance passed with described scenario.

Changed in mos:
status: Fix Committed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.