When running tempest cleanup --prefix, there is an issue with
cleaning up keypairs, because of incorrect listing of them. This
issue is solved by setting the right dictionary key.
Change-Id: I81da575b0f40d2f04440a322b3b694f332c5d58e
Zuul can do better resource allocation if it learns early that a job
will fail even when tests haven't completed yet. This is done by adding
a regular expression that will match on a single failing test within a
tempest run. The regex is copied from what the zuul project has already
been using successfully for some time.
Change-Id: I8c9488a55d04e32384d1523a9658a8711b05c340
The first item of image.disk_formats and image.container_formats
tempest options is used by default if the options are not
specified by the person runnig tempest.
The most likely format to be used and supported are bare and qcow2,
so let's switch to those as default options.
Also, move raw to the second place of disk_formats because
a specific test (test_list_images_param_disk_format from the
tempest.api.image.v2.test_images.ListUserImagesTest class)
requires an image in raw format, but its initializer only
creates a subset of the images (only the first 6 (!) items
of all combinations of container_formats and disk_formats)
which means that raw should be in the first 6 combinations.
Change-Id: Iab3e5205e2ec6cd5a253715ff87853b2805e4819
We are working to get support of new location APIs in glance,
where we are adding hash calculation mechanism once the image
location is added. Since new location api is async,
we need to wait for the task to finish before performing the
cleanup else we will fail due to the existing bug [1].
[1]: https://bugs.launchpad.net/glance/+bug/2045769
Change-Id: Ibde7620c9388fdbc93f741e38c9d63fca44a1bbb
This patch adds a test for verifying that hostname allows
FQDN and can contains periods in hostname starting API microversion 2.94.
This patch also verifies hostname from the metadata api and also checks
hostname allows character length above 64 .
Also, adding a config feature flag dhcp_domain to set the suffix in
order to verify the hostname correctly .
Relates to : https://review.opendev.org/c/openstack/nova/+/869812
Change-Id: If92d4b469bfff4b205801bdd5ea8fd943942b943
These tests can be removed now, because the identity v2 API was removed
from Keystone during Queens cycle and Queens is no longer supported
by current master of tempest.
This also deprecates the related options so that we can drop these
options later.
Change-Id: I0be2273890cecaa26da37624c185c1bd367b2acf
This patch fixes the incomplete cleanup of unit tests for
`record_resources()` function. These unit tests are tied with the
tempest cleanup --resource-list option. The issue was the creation
of None file and resource_list.json file while running unit tests.
Closes-Bug: #2070449
Change-Id: Id8ff96606b30d52ad4c8a6bd31bcb09ae03cc963
`http_image config` option of `image` section is still referring
to old cirros image `cirros-0.3.1-x86_64-uec.tar.gz`.
Updated it to use new cirros image `cirros-0.6.2-x86_64-uec.tar.gz`.
Change-Id: I55e2fbd062c476e839fe8404bae18e918e745bc5
Current http_image config format downloads tar.gz image which is no
more useful since recent changes in glance to fix CVE-2024-32498.
It will fail to import the image using web-download method as we
anticipate disk-format is qcow2 but it's not. Same it will further
fail if image_conversion plugin is enabled in the environment.
Introduced new config parameter http_qcow2_image which will by
default has reference to qcow2 image.
Change-Id: Ic07a55edc9a0ae713265220c9711f3dee48093fc
Recent changes in glance to fix CVE-2024-32498 causes some tests
to fail because those tests assumes disk-format is qcow2 or ami
but it is actually raw and fails to detect the format.
Change-Id: Id92cbcece145426ec95c46649a2f11c3d19ad4bd
This change adds 30mins to the timeout to account
for timeouts on slow nodes and gather more data.
once stabilised we should consider
enabling zswap, or adding more swap to account
for OOM issues thereby allowing a higher concurrency
to reduce overall execution time.
Change-Id: I9529f554acb5b67401ed1cb8700a916c7131ec02
This patch adds a test where we test the following workflow:
1. Create image
2. Create a bootable volume from Image
3. Launch an instance from the bootable volume
4. Take snapshot of the instance -- which creates the volume snapshot
5. Delete the image
In the above workflow, steps 2 and 3 are performed together as we
create the instance as volume backed providing the image ID to
instance create call.
Change-Id: Ie9c27e7e7305147b7081c5f1dcc8f1631081fda5
The Neutron network can have the attribute "external". If a
network is external, can be seen by other projects, same as
when it is shared. However the subnets cannot be seen now.
This issue is being solved in LP#2051831, in [1].
This patch changes the logic implemented in [2], where the
subnets belonging to an external network where hidden to
a non-admin user of other project.
[1]https://review.opendev.org/c/openstack/neutron/+/907313
[2]https://review.opendev.org/c/openstack/tempest/+/484960
Related-Bug: #2051831
Change-Id: I58b2a1ddf9b0fe85d96617863d121aebb7d81e58
When we create mounting paths for devices we have same paths if we create
them on 2 servers. For example for 2 servers and 2 volumes paths look like
this:
- /mnt/vdb
- /mnt/vdc
- /mnt/vdb
- /mnt/vdc
In this patch we add one more directory - id of server to this paths,
that when we have unique patch for each. After this patch paths will look
like this:
- /mnt/226575324/vdb
- /mnt/226575324/vdc
- /mnt/356104023/vdb
- /mnt/356104023/vdc
Change-Id: I7e4b2f68d987cd4d303915062984ce2125667db6
Detect problems related to Glance images not allowing
Glance images to be deleted.
Adds new option:
CONF.volume_feature_enabled.enable_volume_image_dep_tests
Change-Id: Ia216f5be377bf5c35d7addb78d7f8aabb2393405
Change I53b0fa6dcf077191565e2aaf3f08a24b3238a34b added the --slowest
option to the tempest run command for scenario and serial tests but
missed adding it when running api tests.
The --slowest option will show a list of the slowest tests after the
tests finish running.
Change-Id: Ia5a995121748a2c244f060e6ce4b395213bb196e
Currently the single mount path is used for all volume devices, but
this sometimes causes mount to fail because of resource busy. Create
an independent mount path per device to avoid such timing problems.
Change-Id: Ida4381f28fc528ca83b0818aec6c7633112d19fa
In the gate we occasionally see failures like the following:
urllib3.exceptions.ReadTimeoutError:
HTTPSConnectionPool(host='10.209.0.62', port=443):
Read timed out. (read timeout=60)
urllib3.exceptions.ReadTimeoutError:
HTTPSConnectionPool(host='10.209.99.246', port=9696):
Read timed out. (read timeout=60)
where Tempest gives up waiting for an HTTP response after 60 seconds.
It doesn't appear to be related only to one service, as it has been
seen when calling Nova API or Neutron API (the port=9696 example is a
call to Neutron API).
We could try bumping the http_timeout from the default 60 seconds to 90
seconds to see if it helps in slow environments.
Related-Bug: #1999893
Change-Id: I2b2a6a4dee40ee86ca89f6f7f9a83409f8312617
This enables tempest cleanup with record_resources in the tempest-all job
to create a list of all the resources created during the tempest run.
Then the list will be used to run 'tempest cleanup' to clean only
the resources in the list.
Also, enable the dry cleanup with record_resources in tempest-extra-tests
job.
Change-Id: Ia64a1c09d4b5e9815ff5926161d550b56433ec64
This patch adds an option to create purge list that logs every
resource created by Tempest. When used with tempest cleanup
command, it creates a new method of deleting resources created
during one or multiple Tempest runs.
This method solves the problem of accidentally deleting
resources which were not created by Tempest, but by users.
Change-Id: Ide81e6a41799bace211669951b4ceab8635b56ab