yahoo-eng-team team mailing list archive

[Stephen Finucane <sfinucan@xxxxxxxxxx>]

*** This bug is a duplicate of bug 1578155 ***
    https://bugs.launchpad.net/bugs/1578155

This has already been fixed on master and is included in the newton
release. Unfortunately, I don't think we can backport this to Mitaka
(which only accepts high priority fixes now). I'll try, but you may need
to carry this patch yourself (or ask your vendor to do so).

** This bug has been marked a duplicate of bug 1578155
   'hw:cpu_thread_policy=prefer' misbehaviour

-- 
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/1657791

Title:
  cpu pining fails with 1 sibling cpu set and 2 single cpu set

Status in OpenStack Compute (nova):
  In Progress

Bug description:
  Description
  ===========

  CPU pinning fails when the sibling set consists of single VCPUS and a sibling set.
  For example,

  CoercedSet([45]), CoercedSet([31]), CoercedSet([1, 25])

  Considering a case,
  When  a compute with NUMA NODE 1 which has 4 VCPU and 16GB RAM available to use.

  And when you to try to launch a instance with a flavor of 4 VCPU and
  16GB RAM, it is failed to launch i.e "message": "No valid host was
  found. There are not enough hosts available."

  Total Available VCPUS
  1,3,5,7,9,11,13,15,17,19,21,25,27,29,31,33,35,37,39,41,43,45
  Used VCPUS
  3,5,7,9,11,13,15,17,19,21,27,29,33,35,37,39,41,43
  Free VCPUS
  1,25,31,45

  VCPUS STATISTICS
  Total : 22
  Usage : 18
  Free  : 4

  siblings=set([21,45]),set([7,31]),set([15,39]),set([11,35]),set([17,41]),set([3,27]),set([13,37]),set([9,33]),set([19,43]),set([5,29]),set([1,25])])

  1, 25 , 31 and 45 VCPUS are available to use but nova could not able
  to use them.

  As 31 and 45 are not siblings , Nova is not considering them to boot the VM.
  Because of Numa Topology Filter returns with 0 hosts available.

  Steps to reproduce
  ==================

  We have 22 VCPUS available for Pinning in NUMA NODE 1.

  1. Launch an instance with 1 VCPU
  2. Launch an instance with 1 VCPU
  3. Launch an instance with 8 VCPU
  4. Launch an instance with 8 VCPU
  5. Launch an instance with 4 VCPU

  Expected result
  ===============

  Instance should be in active state

  Actual result
  =============

  Instance is in Error state.
  | fault                                | {"message": "No valid host was found. There are not enough hosts available.", "code": 500, "details": "  File \"/usr/lib/python2.7/dist-packages/nova/conductor/manager.py\", line 397, in build_instances |

  Environment
  ===========
  1. Exact version of OpenStack you are running.

     Mitaka Stable

  2. Which hypervisor did you use?
     Libvirt + KVM

  2. Which storage type did you use?
     None

  3. Which networking type did you use?
     Neutron with OpenVSwitch

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1657791/+subscriptions