yahoo-eng-team team mailing list archive

[Matt Riedemann <mriedem.os@xxxxxxxxx>]

Public bug reported:

There are a few places in the scheduler report client that don't account
for shared resource providers, like a shared storage pool.

1.
https://github.com/openstack/nova/blob/09f0795fe0f5d043593f5ae55a6ec5f6298ba5ba/nova/scheduler/client/report.py#L921

That's used in _allocate_for_instance when it compares the current
allocations for the instance vs the allocations that the compute node
thinks it has:

https://github.com/openstack/nova/blob/09f0795fe0f5d043593f5ae55a6ec5f6298ba5ba/nova/scheduler/client/report.py#L929

If those are different, the compute node allocations are going to
overwrite the current allocations for the instance, which could include
a shared storage allocation created by the scheduler. This is
particularly bad since it happens during the update_available_resource
periodic task that happens in the compute service / resource tracker.

2.
https://github.com/openstack/nova/blob/09f0795fe0f5d043593f5ae55a6ec5f6298ba5ba/nova/scheduler/client/report.py#L1024

This is related to #1 and called from the same code in #1, the
_allocate_for_instance method which is the one comparing the current
allocations for the instance to the ones that the compute node things it
needs, which doesn't account for shared resource providers.

** Affects: nova
     Importance: High
         Status: Confirmed


** Tags: compute pike-rc-potential placement resource-tracker

** Summary changed:

- Scheduler report client is not account for shared resource providers
+ Scheduler report client does not account for shared resource providers

-- 
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/1707256

Title:
  Scheduler report client does not account for shared resource providers

Status in OpenStack Compute (nova):
  Confirmed

Bug description:
  There are a few places in the scheduler report client that don't
  account for shared resource providers, like a shared storage pool.

  1.
  https://github.com/openstack/nova/blob/09f0795fe0f5d043593f5ae55a6ec5f6298ba5ba/nova/scheduler/client/report.py#L921

  That's used in _allocate_for_instance when it compares the current
  allocations for the instance vs the allocations that the compute node
  thinks it has:

  https://github.com/openstack/nova/blob/09f0795fe0f5d043593f5ae55a6ec5f6298ba5ba/nova/scheduler/client/report.py#L929

  If those are different, the compute node allocations are going to
  overwrite the current allocations for the instance, which could
  include a shared storage allocation created by the scheduler. This is
  particularly bad since it happens during the update_available_resource
  periodic task that happens in the compute service / resource tracker.

  2.
  https://github.com/openstack/nova/blob/09f0795fe0f5d043593f5ae55a6ec5f6298ba5ba/nova/scheduler/client/report.py#L1024

  This is related to #1 and called from the same code in #1, the
  _allocate_for_instance method which is the one comparing the current
  allocations for the instance to the ones that the compute node things
  it needs, which doesn't account for shared resource providers.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1707256/+subscriptions