In pick_image method to just pick the first image it finds
if it can't find the specific cirros-*uec image.

Also look for the -disk.img since devstack is changing
the default image in:

Id65ebae73b28da7185cb349b714b659af51ef77f

Change-Id: I0e4c38a9e0ae25a922f6b0e09f3f2531f49c88c9
Closes-Bug: 1614118
(cherry picked from commit 01de5a9f68ad474e9bfa213ffeb3501066a52ebd)

Add an other-requirements.txt file containing a cross-platform list
of dependencies needed for running included tox-based tests. Also
include a tox environment for convenience calling the bindep[*]
utility to list any missing system requirements.

This change is self-testing.

For other-requirements.txt see also
http://docs.openstack.org/infra/manual/drivers.html#package-requirements

[*] http://docs.openstack.org/infra/bindep/

Change-Id: I3c62f6924d2d70e23e6787b8da848fbeb0d52ddc
(cherry picked from commit d6ca9e2d7473f6d15a3c76bcd5a7c3cc65b49296)

From API micro version 2.14, evacuate does not need onSharedStorage
flag. This patch check the API micro version and set/ignore the
onSharedStorage flag before call evacuate for each instance.

Closes-Bug: #1581336
Change-Id: I825653d66f94d36a945b8240ec52285827423375
(cherry picked from commit fc90f825f478ac20762c30ec2c84d538acb3ea84)

Change-Id: Ib6edae588603fed5dbe9a2a5eb46f02fe5e78795

WebOb change https://github.com/Pylons/webob/pull/230 changed
the way in which the error response body is formatted such that
it's no longer a nested dict. So we have to handle both the
old convention of an error message key to the response body error
dict and the new way with just the error body dict.

This was reported upstream:

https://github.com/Pylons/webob/issues/235

But given this was apparently implemented as a long-overdue change
in WebOb the behavior is not likely to change.

Change-Id: If653a247d842786d2824b4b3a5c0cde1383ed7ab
Closes-Bug: #1559072

The original tests expect the server console contains a message
  "Uhhuh. NMI received for unknown reason"
but the expectation is wrong because this message is fallback one
when disabling crash dump, that means the test expects the crash
dump is disable. That is not the purpose of this NMI feature.

In addition, this message is output on x86 cpu architecture only
because the message is implemented in
  https://github.com/torvalds/linux/blob/master/arch/x86/kernel/nmi.c#L304
So if we use the other cpu architectures on the test, this test will
fail.

Let's remove the expectation from the tests.

Change-Id: I62fe2e00433fd216def25d99a2f20d17b988fb84
Closes-Bug: #1553597
(cherry picked from commit 471e08a4d41f9d1bcf4703b4a0c64731a94f2510)

Change I01b22593724616bc0a7793c509ecabf095d6927d made the
live_migrate() method in the ServerManager conditional on the
API version requested. This broke the host-evacuate-live
command which is calling ServerManager.live_migrate() directly
with one too many arguments for the v2.25 version of the method.

This updates the host-evacuate-live shell to behave like the
live-migration method and be aware of the API version when
calling the live_migrate() method.

Related to blueprint making-live-migration-api-friendly

Conflicts:
        novaclient/v2/contrib/host_evacuate_live.py

NOTE(mriedem): The conflict is due to
93913c84b67d3aefe79bef106fbe9a83f0bcdb71 not being in stable/mitaka.
Also note that this is not for a blueprint, it's just related
to a blueprint that landed in mitaka but introduced a regression
here when support was added for that blueprint with the 2.25
microversion.

Change-Id: I4dbeb6ebe03f03799b706be2d787d21484b5c664
Closes-Bug: #1561938
(cherry picked from commit 90fbbb29562a905e8f70badf2c31cfb4ec6841ed)