we do have legacy jobs in ussuri and might be in victoria also and
as we are pinning tempest in both stable branches they need to use the
respective stable branches constraints instead of master constraints.
- https://review.opendev.org/q/topic:ussuri-pin-tempest
- https://review.opendev.org/q/topic:victoria-pin-tempest
Hoping this is last time we are updating the tempest setting in d-g
as all jobs in victoria onwards should be migrated to zuulv3 native.
Change-Id: I07e1cb9b038e8d0c8ce83e8554188f9eb72cb339
Without this grenade job in stable/train fails as it uses
master constraints and master is dropping py3.6 support,
like oslo-log already dropped that and is causing issues[2].
Tempest in stable/train is already pinned in devstack[1]
to 26.1.0. So let's use stable constraints for Tempest venv.
[1] https://github.com/openstack/devstack/commit/8a22f738
[2] https://review.opendev.org/c/openstack/neutron/+/841698
Change-Id: I81233ea91cdd24919db245dd8475341d7532daf3
As of Ansible 2.9, using a jinja test as a filter fails; switch it to
"is" format.
The "--sudo" flag has become "--become", switch it.
Change-Id: If75cb934d386ab25240df9f21f9c948843bdb8fb
Stable branches till stable/rocky is using python
version <py3.6. Tempest test those branch in venv
but Tempest tox use the master upper-constraint[1]
which block installation due to dependencies
require >=py3.6. For exmaple, oslo.concurrency 4.0.0
is not compatible for <py3.6.
As we pin Tempest for EM stable brach, we should be
able to use stable constraint for Tempest installation
as well as while running during run-tempest playbook.
tox.ini is hard coded to use master constraint[1] which force
legacy ocata job to use the master constraints and fail
- https://review.opendev.org/#/c/706161/
Devstack can set stable u-c to use via env var but this is
not enough as Tempest recreate the tox env with master u-c
when run test via devstack-gate.
All other branches have been fixed by fixing the run-tempest playbooks
in same way.
- https://review.opendev.org/#/c/705870/
Because stable/ocata and grenade job use devstack-gate, we need to fix this.
This is required to fix the grenade jobs on each stable branch to
get the other fixes in.
[1] bc9fe8eca8/tox.ini (L14)
Change-Id: Ibfc5c3d5e64e274fd674e6c42878d98bb194fa73
Change I1ef1fe564123216b19582262726cdb1078b7650e in DevStack removes
the ability to set the Cinder periodic_interval config option via the
CINDER_PERIODIC_INTERVAL shell var.
The CINDER_PERIODIC_INTERVAL was introduced in DevStack in Havana as
a workaround for Bug #1180976. That bug, however, was fixed in such
a way that it does not require adjusting the periodic_interval in
Cinder, and the workaround was finally removed in Train.
Because devstack-gate is basically frozen as a legacy pre-zuul-v3
supporting infrastructure for DevStack jobs, this patch does not remove
setting the CINDER_PERIODIC_INTERVAL. It simply adds a note explaining
that it's a no-op when using a Train or later DevStack.
Depends-on: https://review.opendev.org/#/c/652711
Change-Id: I7e2cc815ad14036dbae94e73d118b2845ddf5c64
Partial-bug: #1824837
Solves annoying warnings and prevents code from breaking on newer
ansible which dropped --sudo.
The sudo command line option has been deprecated in
favor of the "become" command line arguments. This feature will be removed in
version 2.6. Deprecation warnings can be disabled by setting
deprecation_warnings=False in ansible.cfg.
Change-Id: I0ec17d6bd6a1e3ada366c323ad67e864a7f1b5ed
These seem to be not run for quite some time and they don't
succeed anymore - drop the code to avoid somebody accidentally
running it and wondering. A good example of "if it isn't tested
its broken".
Needed-By: https://review.openstack.org/#/c/581377/
Change-Id: I68d7099e6f5be502f8a578818b5e2e4c82889c15
Adding ETC binaries was a stop gap measure until packages were
available. With the impending release of bionic that'll be soon :)
Also using the mirror on tarballs.o.o was unreliable so these binaries
are cached in the nodepool images as described in[1]. Having this
setting in d-g makes more work than is needed in the near term.
Removing it does introduce a new failure mode if an etcd version is
used in devstack but not yet cacched in the image.
[1] http://lists.openstack.org/pipermail/openstack-infra/2018-March/005871.html
Change-Id: I2a6649ae3094ca65aa31798ee9bddf2fcdd49260
Promote "local hides exit status" (E042) to error and fix issues found
(this is inline with what we do in devstack).
Change-Id: Idd784af066db80bb67792da3dd0aa0ae4c8475db
Modify test-matrix.py so that is can be used as an ansible module.
Add a role in d-g that runs test matrix for the given branch,
role and features file.
Change-Id: Ie36ba0cd7cfcd450b75000a76a64d856f2a83eba
According to codesearch, this variable is not used anywhere. The
key file is likely to be used by the generic switch driver, so
keep that variable but point it at the standard id_rsa key. The
ironic_key is merely another copy of it; it's not clear why it
was a different file to start with.
Change-Id: If67c3fa1d7a5823ee0b0fe29c6902d08f336a299
SSH-based power and management driver interfaces and drivers that use
them have been declared as unsupported in ironic before Newton release,
and are aimed to be removed from ironic code in Pike release.
This patch switches default for DEVSTACK_GATE_IRONIC_DRIVER to pxe_ipmitool,
and removes *_ssh driver from default enabled drivers.
Change-Id: I9b60c9fa24652e9e64e787cd4e5b0152f51e7a28
Depends-On: I74e26943862018b2beecc0b53ba1fb0183a6adf1
Related-Bug: #1570301
Devstack installs tempest into a virtualenv with tox when you enable the
tempest service. Because some devs apparently found this confusing it
also installs tempest globally by default.
This global install is redundant and itself makes the jobs longer but
the extra globaly pacakges affect OSC runtime due to pkg_resources
entrypoint iteration. Just remove it since it is unneeded and safe the
time needed to install tempest and iterate its installation on disk.
Change-Id: Ie4375fa82aa745f961e57bee306b695923617b90
Add SUSE specific tweaks to the various scripts by invoking
zypper instead of yum when needed and adjust various paths
that are SUSE specific accordingly.
Change-Id: I572afb37e9e315e4d829ae0d946715a0fa4e3cf0
For multinode testing we copy content into authorized_keys. Previously
we blew away any old content which meant that if you use eg root to ssh
to the host then you will no longer be able to do so. Address that by
appending to any existing content in authorized_keys files.
Change-Id: I163637d7cc7002b63d6d43eabfa573a56c916d36
In Ibbb430fb1dbf66942168e0cb52d990ab6a2eb8d7 change for devstack,
we download etcd from github for end user scenarios. However this
is not acceptable for our CI. So we need to set the environment
variable for devstack to something in our CI infrastructure. We
need to do this such that it is available to all jobs as etcd
will run for all jobs from now on.
NOTE: We can get rid of the wget in devstack when the distro(s)
package etcd 3.x for general use.
Change-Id: If06731ed540072e78e7fceacf4ebf650ed89f3ea
As described in the new in-line comment, the point of
devstack-early.txt is to capture output from stack.sh before it
switches itself to devstack.txt.
When stack.sh starts, it just does "set +x" ... this means tracing is
coming out on stderr. ansible consumes all out and then, when the
command finishes, prints stdout and *then* stderr. This is really
quite useless for tracing purposes, where obviously you want the
tracing interspaced with any command output.
To avoid this, redirect stderr to stdout on these ansible calls. This
will ensure everything comes out sequentially.
Change-Id: Ieca949407bb371f0ad58ca8d4c9fa692b68e6e87
The all jobs on tempest are supposed to run all tests without a regex
filter. However this was broken by (I really wish I could say
recently) tempest change I3684fce66a799579fa68af119652cafef25a9f03
which switched to tempest run. Since the primary use of the all tox
target there is plugin jobs which almost always run with a regex we
added a '--regex' argument there for backwards compatibility to not
break other projects. This however broke the all job because a --regex
needs an argument. This commit fixes this by adding the "regex" tempest
which will match all tests in the tempest repo and have the same net
effect as not specifying a selection regex at all.
Change-Id: I855a8ec30ed45f0b8227325f76bf48bd9e1e5988
Following I8a0d03ac0296a74e38efd185beb8513866eaf0c4, in light of
saving some Mo of RAM in our test VMs, this turns the Devstack flag
named 'SWIFT_START_ALL_SERVICES' to False.
Note: we can't use the local_conf: stanza in project-config job
definitions because that would require changing many, many job
definitions.
Change-Id: Id249800bff506d8ef7a69c6e342cceefb07b871a
With this flag devstack will enable coredumps for libvirt, which are
saved to /var/core. Compress and save them to the logs.
Change-Id: I381ba3b540c57b2ca5d73087473c8be4b5b32cad
Partial-Bug: 1643911
Depends-On: If7cd54e804a5a389a0d82a325b58f5b41b8ef0db
Add a new flag to run tempest scenarios. This uses the new
tempest scenario tox environment, which run all scenario tests
serially. The DEVSTACK_GATE_TEMPEST_REGEX can be used in
conjunction to filter the list of scenario tests to be executed.
Change-Id: I7d756698a045a436f6d79bbeaf6067879ef6d6dc
This changes how we manipulate the local.conf for the subnodes so that
HOST_IP happens first not last in the local stanza.
Change-Id: I7f0d9bfe53d769e3168ca36d04983f1cd68d9e6d
We created experimental Neutron DVR+HA multinode job with [1].
DVR+HA needs a three node setup with two dvr_snat nodes
and one dvr node. So in this setup, we set
1) controller node's agent_mode to dvr_snat
2) first compute to dvr and other to dvr_snat
[1] https://review.openstack.org/#/c/383742/
Change-Id: I4e08a97158dc4538c6f021c049c6e60fb7293459
This reverts commit 4b227637a8.
This broke neutron gate due to missing .tox/all-plugin/bin/testr in -api job.
Change-Id: Ibde5417d005265d27e939ca29788c21f73b38980
As of I5975c476ae5b26402c209d6e5746e7a5a5a91507 the
variable will no longer be used so we can remove it
after Newton is end of life in devstack.
Depends-On: I5975c476ae5b26402c209d6e5746e7a5a5a91507
Change-Id: I51646812f8fd1cce5709996598f7d0d17e46af45
The current mechanisms for changing local.conf make it so that they
all happen before jobs get running. However projects like neutron do
some rather complicated per job configuration that has to happen on
top of devstack-gate configs. So expose a way for this to merge late.
This does remain a bit error prone because you are unaware of what is
being configured in devstack-gate first. It might be better to make
local.conf configuration either d-g or project oriented.
Change-Id: I53119876b82eeaec19382dbe115cb1740bed0cec
This got lost somewhere along the way, lets add it back. While
openstack-health can give us a view across all runs, it's still useful
to see the slowest tests for a single run.
Change-Id: I75e889192ae0de4e309bef3d740fa206b6cc987e
Tempest removed the deprecated stress tests in
Ib229985ea2a1fee495c9492c9ce1781e6bac1dc6 and their corresponding
gate job in I35acc51392dbc127a451a9899a9183517915381a
This commit cleanup the d-g for the same
Change-Id: Ie082cb3e6d40475404cff33370175ca653715a78
Depends-On: I35acc51392dbc127a451a9899a9183517915381a