Nova compute throws an exception if ephemeral is passed wiithout size.

[Nikola Đipanov]

Alex Xu reports on a related code review https://review.openstack.org/#/c/41647/

"""
Xavier, When create vm as below, i got some error from nova-compute side:

'{"server": {"name": "vm3", "image_ref": "b8cd5faa-a65f-4e47-bfc8-68061574b428", "flavor_ref": "1", "max_count": 1, "min_count": 1, "os-block-device-mapping:block_device_mapping": [{"device_name": "/dev/vdc", "source_type": "blank", "destination_type": "local", "boot_index": 0}], "networks": [{"uuid": "b6ba34f1-5504-4aca-825b-04511c104802"}]}}'

2013-08-30 11:01:32.500 TRACE nova.compute.manager [instance: bc919048-049b-445a-a676-e037d7b1fe31] Traceback (most recent call last): 2013-08-30 11:01:32.500 TRACE nova.compute.manager [instance: bc919048-049b-445a-a676-e037d7b1fe31] File "/opt/stack/nova/nova/compute/manager.py", line 1018, in _build_instance 2013-08-30 11:01:32.500 TRACE nova.compute.manager [instance: bc919048-049b-445a-a676-e037d7b1fe31] set_access_ip=set_access_ip) 2013-08-30 11:01:32.500 TRACE nova.compute.manager [instance: bc919048-049b-445a-a676-e037d7b1fe31] File "/opt/stack/nova/nova/compute/manager.py", line 1392, in _spawn 2013-08-30 11:01:32.500 TRACE nova.compute.manager [instance: bc919048-049b-445a-a676-e037d7b1fe31] LOG.exception(_('Instance failed to spawn'), instance=instance) 2013-08-30 11:01:32.500 TRACE nova.compute.manager [instance: bc919048-049b-445a-a676-e037d7b1fe31] File "/opt/stack/nova/nova/compute/manager.py", line 1388, in _spawn 2013-08-30 11:01:32.500 TRACE nova.compute.manager [instance: bc919048-049b-445a-a676-e037d7b1fe31] block_device_info) 2013-08-30 11:01:32.500 TRACE nova.compute.manager [instance: bc919048-049b-445a-a676-e037d7b1fe31] File "/opt/stack/nova/nova/virt/libvirt/driver.py", line 1689, in spawn 2013-08-30 11:01:32.500 TRACE nova.compute.manager [instance: bc919048-049b-445a-a676-e037d7b1fe31] admin_pass=admin_password) 2013-08-30 11:01:32.500 TRACE nova.compute.manager [instance: bc919048-049b-445a-a676-e037d7b1fe31] File "/opt/stack/nova/nova/virt/libvirt/driver.py", line 2002, in _create_image 2013-08-30 11:01:32.500 TRACE nova.compute.manager [instance: bc919048-049b-445a-a676-e037d7b1fe31] size = eph['size'] * 1024 * 1024 * 1024 2013-08-30 11:01:32.500 TRACE nova.compute.manager [instance: bc919048-049b-445a-a676-e037d7b1fe31] TypeError: unsupported operand type(s) for *: 'NoneType' and 'int'

That because I miss 'volume_size' in the request.
"""

I was also able to reproduce on devstack with latest master.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/44669

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/44669
Committed: http://github.com/openstack/nova/commit/a957beb94720ca3f171149155f9dcabd20e6141b
Submitter: Jenkins
Branch: master

commit a957beb94720ca3f171149155f9dcabd20e6141b
Author: Nikola Dipanov <email address hidden>
Date: Mon Sep 2 11:49:08 2013 +0200

    Interpret BDM None size field as 0 on compute side

    This patch makes sure that even if we receive an ephemeral or swap BDM
    without size information - it is defaulted to 0 by the compute service
    prior to booting the VM, as the size is expected to be present and an
    integer number by the driver code that uses it.

    Closes-bug: #1218861

    Change-Id: Ic6fc15d1cf876aa4fd99e17c582c1ab55e3a18c1