Comment 2 for bug 1268943

The use case is not unit tests, but integration tests: tempest (stress) tests, applications acting as nova client tests.

To better understand the context, it's linked to the bug https://bugs.launchpad.net/nova/+bug/1260771 and its associated gerrit change. An extract from the change is at the end of comment.

When the bug https://bugs.launchpad.net/nova/+bug/1260771 will be corrected, fake + kvm compute nova deployments will be possible. If you use scheduler limiting filters (RamFilter, CoreFilter, etc.) to avoid kvm compute overload, you also constraint fake computes capacity (which is currently hardcoded to a very low value)!

"""
The number of instances supported by fake computes is not limited by physical constraints.
It allows to perform stress tests on a deployment with few resources (typically a laptop).
But you must avoid to use scheduler filters limiting the number of instances per compute (like RamFi
lter, DiskFilter, AggregateCoreFilter),
otherwise they will limit the number of instances per compute.

Fake computes can also be used in multi hypervisor-type deployments in order to take advantage of fake and "real" computes during tests
* create many fake instances for stress tests
* create some "real" instances for functional tests

Fake computes can be used for testing Nova itself but also applications on top of it.
"""