With this patch, a new pytest optional arg has been added:
When it is used, tests are skipped when the regex is found in their test
names. Fully qualified test names are used for this search. Example:
It can be used from CLI with pytest like this:
$ .tox/py3/bin/pytest \
It can be used from CLI with tox like this:
$ PYTEST_ADDOPTS="--skipregex='test_join_chunks_.*_bytes|test_join_chunks_with_unicodes'" \
tox -e py3
It can be used from the Tobiko Infrared plugin by using the following
option (limitation: the skipregex is applied to all the steps from the
the executed Tobiko workflow):
$ infrared tobiko ... --pytest-addopts "--skipregex='test_join_chunks_.*_bytes|test_join_chunks_with_unicodes'"
It can be used from a zuul job by configuring the following variable
within the zuul job yaml file (limitation: the skipregex is applied to
all the steps from the executed Tobiko workflow):
Test test_z99_reboot_controller_galera_main_vip creates a set of 10 VMs.
While the creation is in progress, the test reboots some controller
nodes. Then, it waits until all VMs are created. It uses a timeout
because sometimes not all the VMs reach the ACTIVE status due to the
The timeout value of 4 minutes seems to be not enough in some cases.
This patch increases that value to 5 minutes.
Tobiko tox virtual env creation fails with podman-py 4.8.0 due to
an issue with dependencies.
The new podman-py does not include any feature relevant for Tobiko, so
we can continue using 4.7.0.
This test has the same structure as other faults tests:
- check cloud is healthy
- perform some disruption
- check cloud is recovered and healthy after the disruption
In this case, the disruption is restarting the neutron server on the
overcloud nodes where it is running
When the overcloud nodes use IPv6 addresses, the socket file name
created before this patch was not valid because character ":" cannot be
used on that file.
This patch replaces ":" with "." for the podman socket filename.
The faults test test_network_disruptor_main_vip runs a method to undo
the disruptive action that it executes.
That method is called at the end of the test. If something fails before,
the "undisruption" is not called.
This patch moves network_undisrupt_controller_main_vip to the tearDown
method, which guarantees it would be executed, regardless of passing or
This patch affects only ML2/OVS setups.
Some faults tests fail after reboots or disruptive actions because
the neutron-radvd sidecar containers move from one controller to
another one, which is not really wrong.
Before this patch, the find command could fail due to other reasons
(some directory was not accessible) that are irrelevant to the aim of
this function, which is to find the path of the ovn db ctl files
Depending on the hypervisor/compute were the Ubuntu VM instances are
spawned, their interface names are created following differnt
conventions. Tobiko decides which is the expected name depending on the
OSP version, but that is not a valid method to decide it in some cases.
This patch allows to configure the expected Ubuntu interface name.
On TLS-e environments, manila requests were rejected due to
With this patch, the path to the cerificate is passed to manilaclient
and the requests are answered successfully.
Tobiko does not print any information about the status of this loop
because the sleep_interval is None.
After one hour, the retry loop is broken because the timeout expires,
but there is no information about the reason why it expired.
With this patch, information about the timeout and the retry loop is
printed within the Tobiko logs.
There's a need to create a Manila share resource before the upgrade
process and then validate the share after the process is complete.
To achieve this objective, it is essential to incorporate support for
the Manila client, a task that is accomplished through this commit.
We expect that after traffic to the dedicated controller node(s) is dropped, the controller(s) will be fenced and restarted. Using "check_overcloud_node_uptime" method to check it happens.
The configuration parameter that provides the information about DVR when
ML2/OVN is used is different from the parameter used by ML2/OVS.
1) In case of OVS:
2) In case of OVN:
When DVR is disabled, rebooting or applying network disruptions on the
controllers nodes can affect dataplane traffic because the packets are
not directly routed to the compute nodes.
Due to this, some faults tests fail during the background ping checks.
This patch skips the background ping checks on those tests.
The SharedFixture.cleanUp method is called when setup_fixture fails. The
aim of this is to delete the resources that were not successfully
created and allow next tests to create them properly.
This is not valid with check-resources steps because resources are not
So, in this case, it makes more sense to skip deletion of resources and
run the rest of tests with the existing ones, even if they fail.
With this patch, ResourceFixtures can be declared with
tobiko.required_fixture using kwargs.
Those kwargs can be used to modify default values from attributes used
within the ResourceFixtures subclass constructors
The new ResourceFixture is the base class used to manage the creation of
OSP resources that are not based on Heat Stacks.
The child classes only need to provide some mandatory attributes and
implement methods to create, find and delete resources.
Some tests from some CI jobs fail during check-resources test steps
because the console output is empty and the wait_for_guest_os_ready
method fails due to that.
This package was removed from the OpenStack requirements with  so
keeping it in the doc/requirements.txt file is causing failure in the
As we want to use this theme to build our docs, lets keep it in the
readthedocs_requirements.txt file which isn't checked by the
requirements-check job at all.
This patch also changes docs and releasenotes configs in the way that it
will fallback to one of the default themes provided by sphinx.
Finally it makes some additional changes to the releasenotes conf.py
file to make build of the releasenotes working fine.
Avoid concurrency issues by locking VM creation until it is
Add step that verifies that the VM instance boot is completed before
releasing the lock and before the tests try to ping, ssh, etc that VM.
Add another step that checks the VM FIP is pingable.
Add a third step that checks the VM has completed the cloud-init
Some tests (SameHostNetworkTest and DifferentHostNetworkTest) try to
migrate VMs immediately after creating them in order to move them to the
expected compute node. Sometimes, those VMs fail to contact the metadata
service because their migration happens exactly when the guest OS was
trying to obtain the metadata.
With this patch, those VMs are not migrated until the guest OS is ready.