yahoo-eng-team team mailing list archive

[Armando Migliaccio <1283599@xxxxxxxxxxxxxxxxxx>]

I believe the issue lies in the management for the cache layer between
Nova and Neutron. From what I can tell by looking at this run:

http://logs.openstack.org/02/117902/2/gate/gate-tempest-dsvm-neutron-
full/d5bafe2/logs

I can see that the VM's port is created here:

http://logs.openstack.org/02/117902/2/gate/gate-tempest-dsvm-neutron-
full/d5bafe2/logs/screen-n-cpu.txt.gz#_2014-09-09_05_54_12_795

But then for some reason is not found here:

http://logs.openstack.org/02/117902/2/gate/gate-tempest-dsvm-neutron-
full/d5bafe2/logs/screen-n-cpu.txt.gz#_2014-09-09_05_54_22_403

In fact, I look at the nova code:

https://github.com/openstack/nova/blob/master/nova/compute/manager.py#L4695

It shows that if the for loop does not yield any port, a PortNotFound
exception is raised.

However the port does exist in Neutron, and Neutron does barf
legitimately if one attempts to delete a subnet when ports are still
being allocated.

Looking more deeply, I can see that the cache update is done here:

http://logs.openstack.org/02/117902/2/gate/gate-tempest-dsvm-neutron-
full/d5bafe2/logs/screen-n-cpu.txt.gz#_2014-09-09_05_54_13_124

But subsequently here:

http://logs.openstack.org/02/117902/2/gate/gate-tempest-dsvm-neutron-
full/d5bafe2/logs/screen-n-cpu.txt.gz#_2014-09-09_05_54_13_389

A new update wipes out the port details from _heal_instance_info_cache;
it's possible that the issue here is a dirty read:
_heal_instance_info_cache gets network info, then it is preempted by an
interface event; when it resumes, it overwrites the most up to date info
with the older one. I believe this is a Nova primarily an issue on the
Nova end as well.

** Also affects: nova
   Importance: Undecided
       Status: New

** Changed in: neutron
       Status: Confirmed => Invalid

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/1283599

Title:
  TestNetworkBasicOps occasionally fails to delete resources

Status in OpenStack Neutron (virtual network service):
  Invalid
Status in OpenStack Compute (Nova):
  New
Status in Tempest:
  Invalid

Bug description:
  Network, Subnet and security group appear to be in use when they are deleted.
  Observed in: http://logs.openstack.org/84/75284/3/check/check-tempest-dsvm-neutron-full/d792a7a/logs

  Observed so far with neutron full job only.

To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/1283599/+subscriptions