yahoo-eng-team team mailing list archive

[George Shuklin <1663225@xxxxxxxxxxxxxxxxxx>]

** Project changed: ironic => nova

** Tags added: ironic

** Description changed:

  If nova-compute service is down at the moment of execution 'nova delete'
  for instance, node with this instance will never been cleaned/turned off
  after nova-compute start.
  
  Steps to reproduce:
  1. Prepare ironic/nova
  2. Start instance (nova boot/openstack server create)
  3. Wait until 'ACTIVE' state for instance.
  4. Stop nova-compute
  5. Wait until it become 'down' in 'nova service-list'
  5. Execute 'nova delete' command for instance.
  6. Start nova-compute serivce
  
  Expected result:
  Case 1:
  - Instance stuck in the 'deleting' state until nova-compute is not come back.
  - Node switch to 'cleaning/available' as soon as nova-compute come back
  - Tenant instance (baremetal server) stops to operate as soon as nova-compute is up.
  or
+ Case 2:
  - Instance deleted as usual
  - Node switch to 'cleaning/available' as soon as nova-compute come back
  - Tenant instance (baremetal server) stops to operate as soon as nova-compute is up.
  
  Actual result:
  - Instance deleted as usual.
  - Node has 'active' state with filled in 'Instance UUID' field.
  - Tenant instance (baremetal server) continue to work after nova-compute is up and continue to do so forever (until node is put to 'deleted' state manually by system administrator).
  
  I believe this is very severe bug, because it allows tenants to continue
  to use services regardless of nova report that there are no tenant
  instances running.
  
  Affected version: newton.

** Description changed:

  If nova-compute service is down at the moment of execution 'nova delete'
  for instance, node with this instance will never been cleaned/turned off
  after nova-compute start.
  
  Steps to reproduce:
  1. Prepare ironic/nova
  2. Start instance (nova boot/openstack server create)
  3. Wait until 'ACTIVE' state for instance.
  4. Stop nova-compute
  5. Wait until it become 'down' in 'nova service-list'
  5. Execute 'nova delete' command for instance.
  6. Start nova-compute serivce
  
  Expected result:
  Case 1:
  - Instance stuck in the 'deleting' state until nova-compute is not come back.
  - Node switch to 'cleaning/available' as soon as nova-compute come back
  - Tenant instance (baremetal server) stops to operate as soon as nova-compute is up.
  or
  Case 2:
  - Instance deleted as usual
  - Node switch to 'cleaning/available' as soon as nova-compute come back
  - Tenant instance (baremetal server) stops to operate as soon as nova-compute is up.
  
  Actual result:
  - Instance deleted as usual.
  - Node has 'active' state with filled in 'Instance UUID' field.
  - Tenant instance (baremetal server) continue to work after nova-compute is up and continue to do so forever (until node is put to 'deleted' state manually by system administrator).
  
  I believe this is very severe bug, because it allows tenants to continue
  to use services regardless of nova report that there are no tenant
  instances running.
  
  Affected version: newton.
+ 
+ P.S. Normally nova (with libvirt/kvm driver) would keep instance in
+ 'deleting' state until nova-compute is not come back, and remove it from
+ server (from libvirt). Only after that nova marks instance as deleted in
+ database. Ironic driver should do the same.

** Description changed:

+ 
+ Affected configuration: Ironic installation with Ironic driver for nova.
+ 
  If nova-compute service is down at the moment of execution 'nova delete'
- for instance, node with this instance will never been cleaned/turned off
- after nova-compute start.
+ for instance, baremetal node with this instance will never be
+ cleaned/turned off, even after nova-compute start.
  
  Steps to reproduce:
  1. Prepare ironic/nova
  2. Start instance (nova boot/openstack server create)
  3. Wait until 'ACTIVE' state for instance.
  4. Stop nova-compute
  5. Wait until it become 'down' in 'nova service-list'
  5. Execute 'nova delete' command for instance.
  6. Start nova-compute serivce
  
  Expected result:
  Case 1:
  - Instance stuck in the 'deleting' state until nova-compute is not come back.
  - Node switch to 'cleaning/available' as soon as nova-compute come back
  - Tenant instance (baremetal server) stops to operate as soon as nova-compute is up.
  or
  Case 2:
  - Instance deleted as usual
  - Node switch to 'cleaning/available' as soon as nova-compute come back
  - Tenant instance (baremetal server) stops to operate as soon as nova-compute is up.
  
  Actual result:
  - Instance deleted as usual.
  - Node has 'active' state with filled in 'Instance UUID' field.
  - Tenant instance (baremetal server) continue to work after nova-compute is up and continue to do so forever (until node is put to 'deleted' state manually by system administrator).
  
  I believe this is very severe bug, because it allows tenants to continue
  to use services regardless of nova report that there are no tenant
  instances running.
  
  Affected version: newton.
  
  P.S. Normally nova (with libvirt/kvm driver) would keep instance in
  'deleting' state until nova-compute is not come back, and remove it from
  server (from libvirt). Only after that nova marks instance as deleted in
  database. Ironic driver should do the same.

-- 
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/1663225

Title:
  ironic does not clean or shutdown nodes if nova-compute is down at the
  moment of 'nova delete'

Status in OpenStack Compute (nova):
  New

Bug description:
  
  Affected configuration: Ironic installation with Ironic driver for nova.

  If nova-compute service is down at the moment of execution 'nova
  delete' for instance, baremetal node with this instance will never be
  cleaned/turned off, even after nova-compute start.

  Steps to reproduce:
  1. Prepare ironic/nova
  2. Start instance (nova boot/openstack server create)
  3. Wait until 'ACTIVE' state for instance.
  4. Stop nova-compute
  5. Wait until it become 'down' in 'nova service-list'
  5. Execute 'nova delete' command for instance.
  6. Start nova-compute serivce

  Expected result:
  Case 1:
  - Instance stuck in the 'deleting' state until nova-compute is not come back.
  - Node switch to 'cleaning/available' as soon as nova-compute come back
  - Tenant instance (baremetal server) stops to operate as soon as nova-compute is up.
  or
  Case 2:
  - Instance deleted as usual
  - Node switch to 'cleaning/available' as soon as nova-compute come back
  - Tenant instance (baremetal server) stops to operate as soon as nova-compute is up.

  Actual result:
  - Instance deleted as usual.
  - Node has 'active' state with filled in 'Instance UUID' field.
  - Tenant instance (baremetal server) continue to work after nova-compute is up and continue to do so forever (until node is put to 'deleted' state manually by system administrator).

  I believe this is very severe bug, because it allows tenants to
  continue to use services regardless of nova report that there are no
  tenant instances running.

  Affected version: newton.

  P.S. Normally nova (with libvirt/kvm driver) would keep instance in
  'deleting' state until nova-compute is not come back, and remove it
  from server (from libvirt). Only after that nova marks instance as
  deleted in database. Ironic driver should do the same.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1663225/+subscriptions