yahoo-eng-team team mailing list archive

[OpenStack Infra <1707252@xxxxxxxxxxxxxxxxxx>]

Reviewed:  https://review.openstack.org/490085
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=79c376ec162a0049523f196b3fd10a1585cff1da
Submitter: Jenkins
Branch:    master

commit 79c376ec162a0049523f196b3fd10a1585cff1da
Author: Matt Riedemann <mriedem.os@xxxxxxxxx>
Date:   Wed Aug 2 12:48:01 2017 -0400

    Sum allocations in the scheduler when resizing to the same host
    
    When we resize to the same host, with or without a shared provider,
    the _move_operation_alloc_request code is currently removing the
    compute node resource provider uuid from the new_rp_uuids
    variable since the uuid is the same in both the cur_rp_uuids and
    new_rp_uuids as it's the same compute node provider. So later when
    processing the new destination allocation requests, since new_rp_uuids
    is empty, the new destination allocations aren't part of the request.
    
    This change checks for this case and sums the resource allocation
    amounts when we're resizing to the same compute node.
    
    Change-Id: I1c9442eed850a3eb7ac9871fafcb0ae93ba8117c
    Closes-Bug: #1707252


** Changed in: nova
       Status: In Progress => Fix Released

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to OpenStack Compute (nova).
https://bugs.launchpad.net/bugs/1707252

Title:
  Claims in the scheduler does not account for doubling allocations on
  resize to same host

Status in OpenStack Compute (nova):
  Fix Released

Bug description:
  This code in the scheduler report client is used by the scheduler when
  making allocation requests against a certain instance during a move
  operation:

  https://github.com/openstack/nova/blob/09f0795fe0f5d043593f5ae55a6ec5f6298ba5ba/nova/scheduler/client/report.py#L200

  The idea is to retain the source node allocations in Placement when
  also adding allocations for the destination node.

  However, with the set difference code in there, it does not account
  for resizing an instance to the same host, where the compute node for
  the source and destination are the same.

  We need to double the allocations for resize to same host for a case
  like resizing the instance and the VCPU/MEMORY_MB goes down but the
  DISK_GB goes up.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1707252/+subscriptions