Update CPU topology flavor advice.
The example currently listed would create a two socket CPU with
only one NUMA node, which is a invalid configuration which leads
to undefined results.
Instead list two examples one with a single socket for simpler
end to end tests and a two socket two numa node example for more
advanced low level tests.
The test will prepare hypervisor units in virtual machines by
enabling hugepages configuration and rebooting them and
subsequent required runtime changes. If the hypervisor units
are physical machines the test assumes they already have
received the appropriate configuration from the bare metal
provisioning system.
Launch nested instance using flavor that enables hugepages and
attach it directly to the external network and perform
connectivity tests.
After a successful test the hypervisor units are brought back
to their original state to not influence subsequent tests.
When pgrep_full=True is used, the resulting pgrep can match
more binaries than expected if the search pattern isn't specific
enough. This change makes the search patterns more specific
in order to only match the expected binaries.
Closes-Bug: #1933338
Pyyaml>=6.0 requires to pass the Loader arg to yaml.load(), switching to
yaml.safe_load() recovers the old and expected behavior.
https://github.com/yaml/pyyaml/pull/561
Closes-Bug: #1951650
Do not run the deferred hooks action for neutron gateway as it is
not needed and will end in an action error.
Tested locally here: https://paste.ubuntu.com/p/cSmgxmrTr4/
This closes issue #553
* Add ovn-chassis test
* Yet another refactor to reduce the amount that implementations need to override.
* Add ovn dedicated chassis support
* Fix race with checking wlm
* Add support for running NeutronOvsVsctlTest against any charm and not
only charm-neutron-gateway.
* Add new test case for charm-neutron-gateway and charm-neutron-openvswitch to
verify correct handling of conflicting ext-port and data-port configurations.
Signed-off-by: Przemysław Lal <przemyslaw.lal@canonical.com>
Adjust NeutronNetworkingTest to optionally run tearDown and
re-use existing instances on subsequent runs.
tearDown is controlled through a key under the `tests_options`
dictionary in tests.yaml.
Useful for morphing a deployment and then validating connectivity
for existing instances afterwards.
There are two problems with doing so as part of individual
functional tests:
1) If the application already have the value set the test will
time out waiting for a change that will never be made.
2) python-libjuju ``set_config`` call requires values to be
``str`` regardless of their actual type. Pairing this fact
with the requirement to use the actual type when comparing
values before attempting to set them makes this very confusing
and error prone.
juju/python-libjuju#388openstack-charmers/zaza#348
Use the ``config_change`` helper instead.
When applying configuration to a model the helpers will compare
settings already set to what is being attempted set to be able to
accurately predict model behaviour.
When passing ``bool`` values as ``str`` this does not work and it
may lead to unwanted behaviour depending on the model state at the
time the test runs.
At present we may start interrogating the model for a result of
a change made by the functional test before all units of the
application have started executing.
Commit 55fc718f98 moved a QoS data
plane test from Neutron Gateway tests to the Neutron API tests.
This was probably done because the Neutron API charm gate was
doing end to end tests for all deployment topologies.
On the back of splitting out the data plane tests from Neutron API
gate to the respective data plane charms we need to move this
test back where it belongs.
Apparantly the Neutron Open vSwitch test class have its own
variant of this test already that will be executed by the
neutron-openvswitch charm gate.
Liam Young
Bas de Bruijne
Frode Nordahl
Billy Olsen
coreycb
Felipe Reyes
Martin Kalcok
David Ames
Przemysław Lal
Aurelien Lourot
Dmitrii Shcherbakov