f1c6ac9f2e
It turns out that the ComputeCapabilitiesFilter built-in to Nova doesn't respect capabilities passed in scheduler_hints, so we can't use it for predictable placement. Adding this filter to the undercloud Nova filter list will allow us to do so. Instead of pulling in all of Nova as a test requirement, I've added a fake_nova module to the source tree, which is injected as 'nova' when unit tests are being run. A check is included to make sure nova isn't being imported for real, as well as a README explaining the reasoning behind the fake_nova module. Change-Id: I0618a3b9e3c33af7cdc78db4b6994d463b8aeda9 |
||
---|---|---|
.. | ||
__init__.py | ||
capabilities_filter.py | ||
list.py |