yahoo-eng-team team mailing list archive

[Diko Parvanov <1941892@xxxxxxxxxxxxxxxxxx>]

Public bug reported:

Over-committed source hypervisors shouldn't prevent migration of
shutdown VMs.

Also this error could be be more user friendly and descriptive.

nova-conductor.log showing:

Unable to create allocation for 'VCPU' on resource provider ‘UUID’. The
requested amount would exceed the capacity

Unable to replace resource claim on source host <FQDN> node <FQDN> for
instance

If VCPUs on the source host (openstack host show $FQDN | grep used_now)
are more than allocation_ratio * total cpus (openstack host show $FQDN |
grep total) - then the migration will fail.

Workaround by setting the resource allocation from 1:x to 1:(x+n)
temporary, migrating the instance and then reverting it back to the
original.

for example, if old ratio is 1.0, with a total of 80 physical CPUs:

PROVIDER_UUID=$(openstack resource provider list --name <FQDN> -c uuid
-f value)

Check the current configuration:
openstack resource provider inventory list $PROVIDER_UUID

Change allocation:

CLASS=VCPU
NEW_RATIO=2.0
CURRENT_TOTAL=80
MAX_UNIT=80
openstack resource provider inventory class set --allocation_ratio $NEW_RATIO --total $CURRENT_TOTAL --max_unit $MAX_UNIT $PROVIDER_UUID $CLASS

Perform the migration and change the allocation back to the original
value.

** Affects: nova
     Importance: Undecided
         Status: New

** Summary changed:

- [Informational] Failing migration with 'Unable to create allocation for 'VCPU' on resource provider' and 'Unable to replace resource claim on source host'
+ Failing migration with 'Unable to create allocation for 'VCPU' on resource provider' and 'Unable to replace resource claim on source host'

-- 
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/1941892

Title:
  Failing migration with 'Unable to create allocation for 'VCPU' on
  resource provider' and 'Unable to replace resource claim on source
  host'

Status in OpenStack Compute (nova):
  New

Bug description:
  Over-committed source hypervisors shouldn't prevent migration of
  shutdown VMs.

  Also this error could be be more user friendly and descriptive.

  nova-conductor.log showing:

  Unable to create allocation for 'VCPU' on resource provider ‘UUID’.
  The requested amount would exceed the capacity

  Unable to replace resource claim on source host <FQDN> node <FQDN> for
  instance

  If VCPUs on the source host (openstack host show $FQDN | grep
  used_now) are more than allocation_ratio * total cpus (openstack host
  show $FQDN | grep total) - then the migration will fail.

  Workaround by setting the resource allocation from 1:x to 1:(x+n)
  temporary, migrating the instance and then reverting it back to the
  original.

  for example, if old ratio is 1.0, with a total of 80 physical CPUs:

  PROVIDER_UUID=$(openstack resource provider list --name <FQDN> -c uuid
  -f value)

  Check the current configuration:
  openstack resource provider inventory list $PROVIDER_UUID

  Change allocation:

  CLASS=VCPU
  NEW_RATIO=2.0
  CURRENT_TOTAL=80
  MAX_UNIT=80
  openstack resource provider inventory class set --allocation_ratio $NEW_RATIO --total $CURRENT_TOTAL --max_unit $MAX_UNIT $PROVIDER_UUID $CLASS

  Perform the migration and change the allocation back to the original
  value.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1941892/+subscriptions