can not boot a VM with a custom Aggregates

[lvmxh]

boot a VM with a custom Aggregates will fail

1. Exact version of Nova/OpenStack you are running:
libera

2. Reproduce steps:
step 1. in dashboard, admin -> host aggregates tap
    click "create host aggregate"
         "Host Aggregate Information" tap:
               name: "Pro"
               Availability Zone: "Pro-zone".
          "Manage Hosts within Aggregate" tap
               assign a host to this aggregate.

step 2. create an instance with new Availability Zone "Pro-zone".
      the instance will fail.

3. Relevant log files:
log:
383144-2015-10-06 16:56:54.464 TRACE nova.compute.manager [instance: 6fea2aaa-3d9f-4d88-8987-4764084dbf4d] Traceback (most recent call last):
383145-2015-10-06 16:56:54.464 TRACE nova.compute.manager [instance: 6fea2aaa-3d9f-4d88-8987-4764084dbf4d] File "/opt/stack/nova/nova/compute/manager.py", line 1743, in _prep_block_device
383146-2015-10-06 16:56:54.464 TRACE nova.compute.manager [instance: 6fea2aaa-3d9f-4d88-8987-4764084dbf4d] wait_func=self._await_block_device_map_created)
383147-2015-10-06 16:56:54.464 TRACE nova.compute.manager [instance: 6fea2aaa-3d9f-4d88-8987-4764084dbf4d] File "/opt/stack/nova/nova/virt/block_device.py", line 476, in attach_block_devices
383148-2015-10-06 16:56:54.464 TRACE nova.compute.manager [instance: 6fea2aaa-3d9f-4d88-8987-4764084dbf4d] map(_log_and_attach, block_device_mapping)
383149-2015-10-06 16:56:54.464 TRACE nova.compute.manager [instance: 6fea2aaa-3d9f-4d88-8987-4764084dbf4d] File "/opt/stack/nova/nova/virt/block_device.py", line 474, in _log_and_attach
383150-2015-10-06 16:56:54.464 TRACE nova.compute.manager [instance: 6fea2aaa-3d9f-4d88-8987-4764084dbf4d] bdm.attach(*attach_args, **attach_kwargs)
383151-2015-10-06 16:56:54.464 TRACE nova.compute.manager [instance: 6fea2aaa-3d9f-4d88-8987-4764084dbf4d] File "/opt/stack/nova/nova/virt/block_device.py", line 383, in attach
383152-2015-10-06 16:56:54.464 TRACE nova.compute.manager [instance: 6fea2aaa-3d9f-4d88-8987-4764084dbf4d] availability_zone=av_zone)
383153-2015-10-06 16:56:54.464 TRACE nova.compute.manager [instance: 6fea2aaa-3d9f-4d88-8987-4764084dbf4d] File "/opt/stack/nova/nova/volume/cinder.py", line 388, in create
383154-2015-10-06 16:56:54.464 TRACE nova.compute.manager [instance: 6fea2aaa-3d9f-4d88-8987-4764084dbf4d] raise exception.InvalidInput(reason=e)
383155-2015-10-06 16:56:54.464 TRACE nova.compute.manager [instance: 6fea2aaa-3d9f-4d88-8987-4764084dbf4d] InvalidInput: Invalid input received: Invalid input received: Availability zone 'nova-pro' is invalid. (HTTP 400) (Request-ID: req-daeac6d4-7464-4846-b71f-728e0e3069a1)

[lvmxh (shaohef)]

that's because cinder does not know this new zone.

$ cinder availability-zone-list
+------+-----------+
| Name | Status |
+------+-----------+
| nova | available |
+------+-----------+

but:
$ nova availability-zone-list
+--------------------------+----------------------------------------+
| Name | Status |
+--------------------------+----------------------------------------+
| internal | available |
| |- shhfeng-OptiPlex-9020 | |
| | |- nova-conductor | enabled :-) 2015-10-06T14:41:41.000000 |
| | |- nova-consoleauth | enabled :-) 2015-10-06T14:41:38.000000 |
| | |- nova-scheduler | enabled :-) 2015-10-06T14:41:39.000000 |
| | |- nova-cert | enabled :-) 2015-10-06T14:41:39.000000 |
| nova-pro | available |
| |- shhfeng-OptiPlex-9020 | |
| | |- nova-compute | enabled :-) 2015-10-06T14:41:45.000000 |
| nova-low | available |
| |- stack-OptiPlex-7010 | |
| | |- nova-compute | enabled :-) 2015-10-06T14:41:43.000000 |
+--------------------------+----------------------------------------+

[lvmxh (shaohef)]

resolution:

1. nova should not create an Aggregate with a new zone name.

or
2. nova can create an Aggregate with a new zone name, but it should notify this change to cinder.

[Markus Zoeller (markus_z) (mzoeller)]

@lvmxh:

Could that be a duplicate to bug 1496235 ? This issue got discussed on the ML [1]. The patch for that got merged a few hours ago.

Please double-check if this is the case, until then I put the bug to "incomplete".

[1] openstack-dev; September 2015; "[nova][cinder] how to handle AZ bug 1496235?": http://lists.openstack.org/pipermail/openstack-dev/2015-September/075264.html

[lvmxh (shaohef)]

@Markus Zoeller:

yes, the reason is nova zone is not in cinder zone.

[lvmxh (shaohef)]

but the merged patch, https://review.openstack.org/#/c/227564/4 can not resolve my problem perfectly.

[lvmxh (shaohef)]

@Markus Zoeller:

since nova support create dynamically update zone, why not let user use this good feature?
https://review.openstack.org/#/c/227564/4 just block user to use it.

IMHO, since the nova and cinder are in same cell, they should share AMQP and can get each other’s db info?

The best solution is that when nova zone change, it should notify the cinder.

[Markus Zoeller (markus_z) (mzoeller)]

@lvmxh:

> https://review.openstack.org/#/c/227564/4 just block user to use it.
It prevents the breakage if Nova and Cinder are not in sync regarding availability zones. Which should make the lives of the end users better.

> The best solution is that when nova zone change, it
> should notify the cinder.

On the ML thread [1] it was stated that this would be an orchestration task which is outside of the scope of Nova (IIUC). It seems that this issue will be discussed at the next summit, end of this month. If you like to contribute to this topic, reach out on that ML thread. I close this bug as duplicate to bug 1496235.

[1] openstack-dev; September 2015; "[nova][cinder] how to handle AZ bug 1496235?": http://lists.openstack.org/pipermail/openstack-dev/2015-September/075264.html