This patch will pass only one of the mutually exclusive parameters
'--ntp-heat-env-file' or '--ntp-server' during ceph deploy
Closes-Bug: #2018384
Change-Id: I9c48806f8c0142100aeff79ddb8da0ce3b57cc03
br-tenant is not created as part of standalone
deployments but bridge_mapping was refererring
it. This results into unnecessary Warning[1] in
ovn-controller logs, this patch fixes it.
[1] Bridge 'br-tenant' not found for network 'tenant'
Related-Bug: #1895822
Change-Id: I9b23d6842cd518971b325ffd29b51d171c353b4f
This change removes logic for releases older than train, because these
releases are already EOLed. This covers only standalone roles and
playbooks.
Change-Id: I7b374d5bdb8411a2d773eb1c7ba0e4ded8f5cfbc
Quotation marks should be removed in set statements from
standalone_config jinja template to make the template compatible with
all ansible versions. When quotation marks are used, variable does not
return its value, but is passed as regular string value.
Change-Id: Ia2d664c89a00f139454a4e7b6322eb3fd383e473
While the console is working, it's relying on the "established" state -
if, for any reason, the stream dies, it won't be able to restart
properly.
We already address this case for Undercloud and Overcloud in another
patch:
https://review.opendev.org/c/openstack/tripleo-quickstart-extras/+/857096
The standalone case was overlooked, since it was working just fine.
Change-Id: Idcaaed92e6cc6e5554f1c22d129d962e8daa7272
Now that master is on Quincy but Wallaby is still on Pacific, we need to
make sure the container namespace is the right one. This patch removes
the hardcoded master container namespace and fetch it according to the
openstack branch.
Note that centos9 can be hardcoded because we test both releases on c9.
Change-Id: I7db138fd012e264e87009c524eee02448e793ddf
If the specified default machine type is not
supported on the currently deployed OS, this
patch adds the option to specify a parameter
to override the default.
Related Bug: BZ: 2110535
Change-Id: Id0f0bf0d9efb2136ae49b8475e80774dd0578230
By adding new ceph jobs that configure their own container
namespaces to run tests against different image tags, we started to
face some issues in standalone role, which has some conflicting
conditions on setting custom ceph parameters.
This patch tries to fix these conditions to make them easier to
use.
Change-Id: If108dc66f15333b7ba90fff980f320e06a21bbff
I982dedb53582fbd76391165c3ca72954c129b84a introduced a task
to update registries.conf to trust the docker_registry_host
with the Ceph container. The update is only made in v2 format.
The podman pull works on el9 but because el8's registries.conf
has v1 format sections by default the v1/v2 mix causes 1980869.
As described in Ic35155f04bf05913b9e9b8eaa22fe6c02515396c, which
adds v2 support, tripleo "nuke(s) the existing configuration" and
this patch adds a task to do the same. Why not just run the role
from the same change?
The tasks from Ic35155f04bf05913b9e9b8eaa22fe6c02515396c can be
run by the tripleo.operator role tripleo_ceph_deploy by passing
tripleo_ceph_deploy_skip_container_registry_config=false, but
this option is true in order to be compatible with the changes
in containers.yml which includes configuration for local mirrors.
This is a scenario unique to our tesing environment. It could be
refactored but this patch gets CI working on el8 (it had originally
only been tested on el9).
Change-Id: I077f1a317e43ba29da20e74955940b585c201502
Closes-Bug: #1980869
Deploy Ceph before the overcloud when testing Wallaby.
Follow up to I982dedb53582fbd76391165c3ca72954c129b84a
Change-Id: I8ec7d50ba2d5d9fab9343ca0a4707ddc09fc7797
If the standalone job is periodic, use 'openstack overcloud
ceph deploy --container-namespace' to pull the container
directly from quay.rdoproject.org.
Standalone scenarios usually consume containers from a
content-provider but periodic jobs pull them directly
from quay.rdoproject.org. The periodic jobs then push
them to the undercloud as a container registry. Non-
periodic jobs do not push to the undercloud registry.
When I982dedb53582fbd76391165c3ca72954c129b84a merged,
periodic standalone jobs broke because the undercloud
container registry was not configured when 'openstack
overcloud ceph deploy' was run. Because the container
prepare file has a push_destination, deployed ceph
assumes the containers were prepared in advance so it
swaps out the container namespace as indicated by
push_destination directive. Though this still happens
we override it again with --container-namespace.
Change-Id: I1abfbbd23ca93c01393d05057806ba9cc846fbed
Closes-bug: #1978998
I982dedb53582fbd76391165c3ca72954c129b84a added
the ceph deploy task before the overcloud.
This action was also happening on stable branch
jobs in non-branched repos.
This review adds a condition to skip the task
for current branch_override jobs.
Depends-On: https://review.opendev.org/c/openstack/tripleo-common/+/846287
Closes-Bug: #1978956
Change-Id: If2c9fc02c62a0cf060a73938ca8f8b4101d8c640
Add new tasks file to the standalone role so that CI can
deploy Ceph before the Overcloud. Tasks file uses three
tripleo-operator-ansible roles which create a Ceph spec,
create the ceph-admin user, and deploy Ceph by calling
tripleo-client commands.
Additional tasks file storage-network.yml creates a
dummy interface for a separate storage network so that
Ceph has a network to bootstrap in before the control
plane network (192.168.24.0/24) is created by standalone
overcloud deployment. Same tasks file creates Heat inputs
so overcloud is deployed with network isolation in order
to use the storage network.
Contains a TODO item so that this scenario will also
be tested for W after supporting patches are backported.
Change-Id: I982dedb53582fbd76391165c3ca72954c129b84a
This patch updates extra-commons to check if a ceph namespace, tag or
image was provided within jobs definition.
Change-Id: Ibae7b900cf027c6b610c55399aad3d56e8fdb512
Signed-off-by: Douglas Viroel <dviroel@redhat.com>
To be able to run properly dns_domain_name related tests
from the neutron_tempest_plugin.api.test_ports module,
Neutron has to have configured non default dns_domain_name option.
This patch set dns_domain_name option to "openstackgate.local"
which is the same value like configured in all of the Neutron CI
jobs by the Neutron Devstack plugin.
We have enabled this test in 16.2 by cherry-picking [2] and [3]
in downstream. Same test should not be triggered in train/ussuri
as tempest is pinned - So adding dns_domain_name should ideally
fix downstream(and add coverage) without affecting train/ussuri.
Couldn't find any reason in related-bug on why NeutronDnsDomain
cannot be added for train/ussuri version.
Related-Bug: #1950815
[1] https://bugzilla.redhat.com/show_bug.cgi?id=2053378
[2] https://review.opendev.org/c/openstack/neutron-tempest-plugin/+/821079/
[3] https://review.opendev.org/c/openstack/neutron-tempest-plugin/+/820456/
Change-Id: I3cd8d0bcc01021b27ca7dc3a3670e3fce60bf861
Commit I88b6570d5593e301fe60842bc1dceff22901e037
had some objections raised as to not override the
default NTPserver.
This patch removes the defaults and only specifies the
NTP server if it is defined in the job.
Change-Id: I51ede18650daad46048b672476696fe6ab075d84
New qemu-kvm-6.1.0 version is buggy version and excluded in master
and wallaby (Ia6a9e01ca2adbde1e7a0b7cce9fe5842a0d54b1b,
Iac4425d655df2f519f5b467e3754dad08ac9b1f5).
New version was impacted q35 machines for CentOS-8 and Stream-9.
qemu-kvm-6.1.0 + NovaLibvirtNumPciePorts: 12 + q35 instances
for CentOS8 and Stream works fine.
Related-Bug: #1950916
Signed-off-by: Amol Kahat <amolkahat@gmail.com>
Change-Id: I39c00edd97e321963d1ced74dcf5a2a27fa032cc
NtpServer should be able to be defined in the
standalone config and allowed to be overridden
per environment via zuul parameters.
Change-Id: I88b6570d5593e301fe60842bc1dceff22901e037
Related-Bug: #1955508
There are two issues with the current configuration.
1) clock.redhat.com is not publicly accessible so we need an
alternative.
2) the line server setting is being configured as
"['clock.redhat.com']" which is not a valid server
It should be noted that this hackery of a CI environment file from THT
is not a good solution. This bit should get fixed so these values end up
in the standalone parameters rather than editing
THT/ci/environments/standalone-ipa.yaml.
Closes-Bug: #1955508
Change-Id: If2cbcab21b3a5a3826cf076205dd3222f5893078
To be able to run properly dns_domain_name related tests
from the neutron_tempest_plugin.api.test_ports module,
Neutron has to have configured non default dns_domain_name option.
This patch set dns_domain_name option to "openstackgate.local"
which is the same value like configured in all of the Neutron CI
jobs by the Neutron Devstack plugin.
Related-Bug: #1950815
Signed-off-by: Amol Kahat <amolkahat@gmail.com>
Change-Id: I5a0188524ee05053c708ceb5c564bb3ebcb614dc
This is workaround for downstream osp17 jobs, TripleO now defaults to
libvirt's "modular daemons" after patch[1]. But this required libvirt > 7.4,
this is a temporary workaround to move back to monolithic modular libvirt daemon
till we move RHOSP17 to RHEL9 which will provide libvirt > 7.4.
[1] https://review.opendev.org/c/openstack/tripleo-heat-templates/+/814568
Co-Authored-By: Sandeep Yadav <sandyada@redhat.com>
Signed-off-by: Soniya Vyas <svyas@redhat.com>
Change-Id: I32a87e966625e44f5bd32992271aad4ddb753a9d
Set DockerInsecureRegistryAddress consistantly
across standalone deploy/upgrade and undercloud
deploy roles.
Also include following changes:-
- Use local_docker_registry_host in place of standalone_ip
as that's common for both undercloud and standalone deploys.
- Use quay.io instead of docker.io as a default registry due
to rate limits with docker.io.
Follow up of [1].
[1] https://review.opendev.org/c/openstack/tripleo-quickstart-extras/+/814699
Change-Id: I4286e1d0867fc972517f8e27665f43777c2da52a
Temporarily add a workaround for EL9 by setting
~~~
NovaLibvirtNumPciePorts: 12
~~~
It's basically a workaround for a bug [1] in qemu that is currently
being worked on.
[1] https://bugzilla.redhat.com/show_bug.cgi?id=2006409
Related-Bug: #1947861
Change-Id: I42d67f9f4bdf05b5dafb7169c7e030b3e340e6c9
When adding repos to update in standalone
and undercloud, only add a comma before the list
of additional repos if a previous update repos
exist.
Change-Id: Ic1568c99e16eecd607f063fe8a95fa8c4aa9ed2b
This reverts commit 7317b44130.
Reason for revert:
No longer required since change: I79d141358ead2a2c90fdb728e17926d96d498a28 we now do it in TripleO client.
Change-Id: I2b7f00b8c9b56ba38c271de90fadac3fdd85fcfe
TripleO deploy archives output directory, and since
we are using home directory as output directory, it
archives everything in home(like zuul cloned repos,
virtual envs, pip cache etc) making logs huge.
This patch changes output directory to <working_dir>/tripleo-deploy
which is already collected in logs[1] to ensure only
the required part is archived and collected.
[1] https://review.opendev.org/c/openstack/tripleo-ci/+/784075
Closes-Bug: #1927307
Change-Id: I62565c75e981ebfb38d5a3566f111a6c07f5b5b6
atm there is an option to turn on local container
builds for standalone that is defined in zuul only.
ATM we are considering turning this on for all standalone
master jobs to reduce the load on docker.io
Change-Id: Iad71c1af41b2039591eccdb420c2a35b05ade4ac
This change updates the default bridge mapping from datacentre:br-ex to
datacentre:br-ctlplane. We're doing this because in the standalone in
CI, we configure a br-ex before running the standalone (via
undercloud-setup) and want to attach our br-ctlplane to it. We then want
to ensure that we use br-ctlplane for the neutron access to the external
network to prevent weird routing issues when we have two bridges on the
same subnet.
Depends-On: https://review.opendev.org/#/c/757605/
Change-Id: I0e5aa3f58746dc0b92bd35ade7792f323b5647f7
Related-Bug: #1895822
Wires up the content provider container registry address into
DockerInsecureRegistryAddress. Makes sure we are using the
_address_branch version in all other references to the
job.registry_ip_address var.
Needed by the upgrade jobs in https://review.opendev.org/753403
Change-Id: If9b516860ad3f102e39cb8755cd8aee6f9aae413
When provider_job is set to true, it will pull the container
images from quay.io and will push it to local
registry using non_tripleo_containers.yml.
When consumer_job is set to true in child job, then
custom_ceph_parameters will be setted to true and ceph containers
related namespace will be changed to local registry
to pull the ceph containers so that container built and pushed
to local registry in provider job, will be consumed here.
It also adds standalone job which depends on content provider
job and updates the ceph tag.
Setting +standalone_container_ceph_updates to false as with in
tasks it is setted default to false.
Moved local registry host and port to extra-commons role
to easily share between roles.
It also fix the non_tripleo_containers playbook to exclude kolla
containers for train release.
Note to self:
tripleo-standalone-scenarios-pipeline will be added later
once standalone-upgrade job is ported to centent provider.
Change-Id: Ie3f7bafe3b6ee8a7879ebb3d1d5b47b20181b452
Signed-off-by: Chandan Kumar (raukadah) <chkumar@redhat.com>
Co-authored-by: Alex Schultz <aschultz@redhat.com>