We're more likely to time out while waiting for the
action's result if we launch an action while the
model is still executing. Thus it's safer to wait
for the model to settle between actions.
* Check blocked ovn-chassis units with bad config
LP bug#1919481 [0] needs to have a func-test to check if units can
handle bad bridge-interface-mappings configuration and set units
workload to block state and return to active state with good config.
[0] https://bugs.launchpad.net/charm-ovn-chassis/+bug/1919481
* Changing test_wrong_bridge_config to use config_change
Instead of using block_until_unit_wl_status, change the test_config
to be able to check blocked state for ovn-chassis units
Co-authored-by: Aurelien Lourot <aurelien.lourot@canonical.com>
After zaza#451 Zaza's application readiness detection
has been improved, which now exposes issues from the
past that were hidden. In other words, we were relying
on a bug on Zaza's side and need to get this straight.
In this change, a charm test that subclasses the ManilaBaseTest
class can also opt-in to re-validating a share after restarting
the share services by overriding the _restart_share_instance
method.
When running Trilio tests in a scenario test there is no 'charm_name'
so the application name needs to be explicitly set when calling
the setup class otherwise some config, such as lead_unit, is
missing.
Starting with OpenStack Wallaby the LBAASv2 API will require
authentication with system scoped token unless the user has
the 'load-balancer_admin' role.
Grant role on test setup, and remove it again on clean-up.
Starting with OpenStack Wallaby the LBAASv2 API will require
authentication with system scoped token unless the user has
the 'load-balancer_admin' role.
Grant role on test setup, and remove it again on clean-up.
When Vault is running in a LXD container, test_vault_reload is expected
to fail because mlock cannot be enabled in LXD containers. Skip it when
running in containers is detected. It's useful for local development
environments with Juju's localhost/LXD provider. The test won't be
skipped in the OpenStack Charm CI since the Vault units will be run in
KVM, not in LXD containers.
Vault's runs its tests twice. In some circumstances the removal of the
vault key from the test_config dictionary can throw a KeyError. Do not
fail when this happens.
The simplestreams service relies on the Keystone Service catalog.
To help debug any issues occurring during test execution, log the
contents of the Keystone service catalog in the event of failure.
Related-Bug: #1930654
Set up Ceph-Radosgw in the same model as an external S3 backend for
Glance:
- create an S3 user in Radosgw
- create an S3 bucket for Glance
- inject those parameters into Glance charm configurations as an
external S3 backend
- upload an image to the S3 backend via Glance, and check the size of
the image in the S3 bucket directly
Co-authored-by: Connor Chamberlain <connor.chamberlain@canonical.com>
Launchpad: #1911011
Gerrit review: https://review.opendev.org/c/openstack/charm-nova-compute/+/783371
* Functional tests for `instance-count` action on nova-compute charm.
Test for `remove_from_cloud_actions` waits at the end until
the nova-compute service is available again. However there's a few
seconds lag until that host shows up also in `hypervisor-list` which
can cause race conditions with other tests.
