python-openstackclient/releasenotes/notes/bug-1647406-c936581034a1b6e4.yaml
Nikita Gerasimov 95c8661f86 Switch server create to block_device_mapping_v2
Current compute_client.servers.create() relies on block_device_mapping
arg which is legacy[1]. "block_device_mapping" format require
device_name which is leads to hard-coded hack in --volume key handler
to KVM specific. "block_device_mapping_v2" format is more friendly
to hypervisiors.
Support of block_device_mapping_v2 appear in python-novaclient 2.16.0,
openstackclient require at least 2.29.0

Makes options --volume and --block-device-mapping work simultaneously.
Appends --block-device-mapping data even if --volume used.
After bug 1383338 only --volume was taken when both are used.

[1]http://docs.openstack.org/developer/nova/block_device_mapping.html

NOTE(dtroyer): I moved the new test_boot_from_volume() functional
test to Ie51b1c375c5940856ec76a5770df3c6bd18a3eba to test our
previous behaviour.  The only changes required to support the new
behaviour should be that the empty_volume is now attached in that test.

Change-Id: I7bac3d870dd9ca404093142f8bce22a62e49180d
Closes-Bug: 1647406
Closes-Bug: 1497845
2017-01-26 02:02:30 +00:00

17 lines
788 B
YAML

---
fixes:
- |
Allow ``--block-device-mapping`` option to work correctly with
``--volume`` option in ``server create`` command.
After :lpbug:`1383338` ``--block-device-mapping`` was ignored if
``--volume`` was present. Block device mappings are now appended
to the mapping created by the ``--volume`` option if it is present.
The device name of the boot volume specificed in the ``--volume`` option
is no longer assumed to be *'vda'* but now uses the hypervisor's boot
index to obtain the device name. This maintains the status quo for
**QEMU/KVM** hypervisors but **XEN**, **parallels** and others
*virt types* that have device naming is different from ``vd*``
should now also work correctly.
[:lpbug:`1497845`]
[:lpbug:`1647406`]