tempest branch doesn't point to its master

[Xin]

Gating for OSA master kept failing on:
"==============================", "Failed 1 tests - output below:", "==============================", "", "tempest.api.identity.admin.v3.test_roles.RolesV3TestJSON.test_list_all_implied_roles[id-3748c316-c18f-4b08-997b-c60567bc6235]", "-----------------------------------------------------------------------------------------------------------------------------", "", "Captured traceback:", "~~~~~~~~~~~~~~~~~~~", " Traceback (most recent call last):", " File \"/openstack/venvs/tempest-master/lib/python2.7/site-packages/tempest/api/identity/admin/v3/test_roles.py\", line 393, in test_list_all_implied_roles", " self.assertEqual(2, len(sorted_rules))"
...

While the tempest master code was updated two month ago with commit:
https://github.com/openstack/tempest/commit/7d2b636a30057ed8db8cfd4fe2248f509b3570f1

The current test template on OSA which makes tempest choosing its branch:
https://github.com/openstack/openstack-ansible/blob/master/tests/roles/bootstrap-host/templates/user_variables.aio.yml.j2#L216

Problem was confirmed happening on OSA(SHA:2da59eeabc4b064e53d73a732a9a2c2908d3068d) so far.

[Jean-Philippe Evrard (jean-philippe-evrard)]

Tempest should not point to master. This was implemented as a temporary workaround for master branch, until we discussed further what needs to be done.

IMO, we should always have master pointing to tempest master, and stable branches using pypi.

That can be implemented by being selective, add "tempest*" entries in a file like: https://github.com/openstack/openstack-ansible/blob/master/playbooks/defaults/repo_packages/openstack_services.yml and removing them in stable/rocky and stable/queens branch.

[Jean-Philippe Evrard (jean-philippe-evrard)]

And removing previous patch of OSA, ofc.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/590020

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/590778

[OpenStack Infra (hudson-openstack)]

Change abandoned by Frank Zhang (<email address hidden>) on branch: master
Review: https://review.openstack.org/590020
Reason: Since https://review.openstack.org/#/c/590778 is including more details and more correct change.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/rocky
Review: https://review.openstack.org/591357

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/591357
Committed: https://git.openstack.org/cgit/openstack/openstack-ansible/commit/?id=5c12f15b6c94ea7923427bd4388a7f5e5172936a
Submitter: Zuul
Branch: stable/rocky

commit 5c12f15b6c94ea7923427bd4388a7f5e5172936a
Author: Jean-Philippe Evrard <email address hidden>
Date: Fri Aug 10 16:02:18 2018 +0200

    Temporarily use tempest master for all builds

    A fix [1] was introduced in openstack-ansible to use tempest from
    the master branch in the AIO, but respecting the installation
    from PyPI everywhere else. This allowed users to chose
    the appropriate behaviour for their use cases.

    This patch [1] also made sure we didn't have anything else to do
    when the patch would land on stable branches, as this would
    automatically get disabled.

    However, as of today, there is still no new version of tempest,
    and our stable/rocky branch has been cut. Therefore the fix that
    was in place is useless and another approach needs to be
    taken.

    We will move to use tempest master for master and stable/rocky
    branch until the fix that caused this series of patch to
    happen is released [2] for tempest on PyPI.

    This patch is required for the RC1 freeze on rocky, however
    this is blocked due to a master patch in ceph-ansible.
    This should unblock the situation temporarily by using
    the sha for ceph-ansible that is planned in the RC1 freeze.

    [1]: https://github.com/openstack/openstack-ansible/commit/abe0b22c767444408a10c0c210aa3b70f319dc95
    [2]: https://github.com/openstack/tempest/commit/7d2b636a30057ed8db8cfd4fe2248f509b3570f1#diff-5c9acbc10dc9d27b47985cd74ab100f6

    Change-Id: I42e0cefb28184964a50f66b18d8eba6d2a04eca7
    Closes-Bug: 1786095

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/590778
Committed: https://git.openstack.org/cgit/openstack/openstack-ansible/commit/?id=80dfbe783deae48979c1c0093d4b03fc99029e15
Submitter: Zuul
Branch: master

commit 80dfbe783deae48979c1c0093d4b03fc99029e15
Author: Jean-Philippe Evrard <email address hidden>
Date: Fri Aug 10 16:02:18 2018 +0200

    Temporarily use tempest master for all builds and fix configs

    A fix [1] was introduced in openstack-ansible to use tempest from
    the master branch in the AIO, but respecting the installation
    from PyPI everywhere else. This allowed users to chose
    the appropriate behaviour for their use cases.

    This patch [1] also made sure we didn't have anything else to do
    when the patch would land on stable branches, as this would
    automatically get disabled.

    However, as of today, there is still no new version of tempest,
    and our stable/rocky branch has been cut. Therefore the fix that
    was in place is useless and another approach needs to be
    taken.

    We will move to use tempest master for master and stable/rocky
    branch until the fix that caused this series of patch to
    happen is released [2] for tempest on PyPI.

    A temporary workaround was also added to ensure ceph can still
    be built, by freezing the ceph-ansible role sha.

    The checksum fill option, was useful a long time ago, but at this point
    its just some lovely technical debt that we carry around like an old
    dirty security blanket. This change sets us free from the sins of our
    past.

    We seem to have a lot of jobs that timeout but seem to be very
    close to completing, so let's bump the timeout by 20 minutes.

    To make up for it, we drop most of our non-voting jobs because
    they are taking up resources and always failing. This should
    decrease the amount of resources we're consuming.

    Also, instances are failing to launch when jobs hit any nodepool
    VMs inside OVH.

    [1]: https://github.com/openstack/openstack-ansible/commit/abe0b22c767444408a10c0c210aa3b70f319dc95
    [2]: https://github.com/openstack/tempest/commit/7d2b636a30057ed8db8cfd4fe2248f509b3570f1#diff-5c9acbc10dc9d27b47985cd74ab100f6

    Change-Id: I42e0cefb28184964a50f66b18d8eba6d2a04eca7
    Closes-Bug: 1786095
    Signed-off-by: Kevin Carter <email address hidden>

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/openstack-ansible 18.0.0.0rc2 release candidate.

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/openstack-ansible 19.0.0.0b1 development milestone.