Add steps to multinode-overcloud-update.yml playbook which are required
before running full minor update on overcloud. The prerequisites for
starting overcloud update for tripleo-upgrade are:
- updated dlrn repos definitions on overcloud nodes
- updated containers-prepare-parameter.yaml file on undercloud
Resolves: #2015993
Depends-On: https://review.opendev.org/c/openstack/tripleo-ci/+/880162
Depends-On: https://review.opendev.org/c/openstack/tripleo-quickstart/+/883307
Change-Id: I2829307d1013e96317338eccc7a4b16376fb1d2b
Removes the unused multinode-overcloud-mixed-os-deploy-compute.yml
and renames the -control.yml to just -deploy.yml. Patch is needed
by [1][2]
[1] I04c88580de952c56e219dbf76732d30f2fae3979
[2] Idfa6653b33cf8bd90b5dfe1b81573670bfc5b075
Change-Id: Iec88d20b83aba4f4832a3b31b42d3780188ab2fd
This patch looks huge, so let's split things in the following listing:
- rename role to match the naming convention
- role name shouldn't get hyphen, so let's use underscores instead.
Playbooks don't need to follow this rule.
This should make the role more portable in the future.
- modify role
- "disable" the main.yml - we must call specific task file.
- no more condioned deploy based on variables. Makes it easier to
follow and maintain.
- update how the edpm inventory is edited
- use plain YAML
- output plain YAML
- this format is consistent with latest changes in tripleo-ansible[1]
These changes make the overall role easier to understand and maintain,
while ensuring the playbook is also easier to read. It may also make the
deploy a bit faster with the new way to run the various tasks.
- update playbook
- boostrap both nodes in a 2-stages way (check comment as to why).
- deploy compute from the undercloud (mimic what an operator would do)
- validate the compute from the undercloud (no actual change here)
- target specific task_file during the different stages. It's easier
to understand what's going on in there
These changes should allow a better understanding of the overall playbook.
- update edpm deploy script to use an environment file instead of CLI
params
- this makes the script lighter and easier to understand
- adding new parameters in the env file is easier
- reading the env file for debugging is better
[1] https://review.opendev.org/c/openstack/tripleo-ansible/+/869357
Change-Id: I263f424a5a08214b9c46c5ea73b6b428709ad47b
Remove auto-reverse from IPA server install as
this option is causing errors in environments
where the zone being created already exists and
is owned by some other dns server.
See related change in:
https://review.opendev.org/c/openstack/tripleo-quickstart-extras/+/737058/.
This patch also stops the unbound service if it
is running on the undercloud.
Change-Id: Ie46809379e5c143910560b4d3f7cb0eee6f80ea1
It adds validate_edpm.yml playbook to list
external compute on the controller node.
Signed-off-by: Chandan Kumar (raukadah) <chkumar@redhat.com>
Change-Id: I5e0b16bf34075d663a52756c4996512610927e47
This role is no longer used, and was a long ago replaced with tempest.
Right now, there are still some pieces of code that overwrite the
tempest testrepository.subunit with the pingtest, and so, the best
approach right now is ensure that the validate-simple is not being
called anywhere.
Change-Id: I5a2993fa93c08f842e936c80b0adf4f9fb5c3b0f
This change removes logic for releases older than train, because these
releases are already EOLed. This covers only tempest roles and
playbooks.
Change-Id: I0b50cad30107a83b562b8c22b2afd7af7636c615
This change removes logic for releases older than train, because these
releases are already EOLed. This covers only undercloud-* roles and
playbooks.
Note that this change also removes the containerized_* variables
because current active releases(>=Train) no longer supports
non-containerized deployment. The containerized_undercloud_upgrade
variable is still left because the variable is still required by
the tripleo-ci-base-multinode-standard job template.
In addition, TripleO UI support is removed because the component was
retired before Train.
Change-Id: I5132e5bac4ddaa6f76564cab67e486cb5a4e2bbb
This change removes logic for releases older than train, because these
releases are already EOLed. This covers only overcloud-* roles and
playbooks and the other items will be covered by separate changes.
Note that this change also removes the containerized_* variables
because current active releases(>=Train) no longer supports
non-containerized deployment. The containerized_overcloud_upgrade
variable is still left because the variable is still required by
the tripleo-ci-base-multinode-standard job template.
Change-Id: If29ec2c2219a28a1f79db0e552e2c622c0a7bda6
external-data-plane-management role is a new role to deploy
standalone compute node by using tripleo-ansible roles.
This role includes will run on two nodes and does following
things:
- Bootstrap first node and deploy standalone
- Generate 99 standalone vars and copy it to second node
- Bootstrap second node and install packages
- Copy 99 standalone vars to tripleo-inventory
- Deploy compute serices.
It will be used in tripleo-external-compute-deployment-pipeline
job template[1]
[1]. https://review.opendev.org/c/openstack/tripleo-ci/+/843836
Signed-off-by: Chandan Kumar (raukadah) <chkumar@redhat.com>
Change-Id: I476c7299dec209c89db87e2182a35ec6a4924f92
dstat logs were running in excess of 13G and overrunning
the space on baremetal boxes. dstat collection is not
needed on the actual host boxes.
Change-Id: I418907d26c160ccceff53edb8332e277d88214ca
In the periodic jobs we don't have a content provider to draw the
container build id from for the compute node container image
prepare. Adds a conditional task to resolve tripleo-ci-testing.
Change-Id: Ic01b5bba246c96f3f44679aab3be724c3c5fdf35
This adds some new playbooks to be used by the new multinode mixed
os job (See topic at [1] for related reviews). Also updates the
extras-common and undercloud-deploy roles to allow the mixed-os
job to use the correct content provider.
[1] https://review.opendev.org/q/topic:oooci_mixed_rhel
Change-Id: Ie7a95056bfb0a90d2e6fb6e5ec9c8ed66f8ef31e
Since we weren't using the tempest-allow role before, the logic was set
to be executed when tempest_test_whitelist wasn't defined, which is
fine, but to run skipped jobs, the allowed list was being overwriten.
This patch add the logic to only run the tempest-allow role when
tempest_run_skipped is set to false (default).
Change-Id: Ie4a5b6a764d19ce49f703804daaa50425be1bb6f
With this patch, we are adding a new role discover-latest-image to
discover the latest Centos Stream image, We are limiting the scope
for C9 for now, We can extend the role for Centos 8 image discovery
in future if there is a need.
Also, we are updating baremetal-prep-virthost.yml playbook to use
this new role for c9 case and in combination with patch[1] - we will
get rid of undercloud baseos hardcoding and use the latest discovered
image.
[1] https://review.opendev.org/c/openstack/tripleo-quickstart/+/838990/
Change-Id: I3a97edc63e96725a1690335f557508d56ed10258
It's make sure, all the overcloud nodes are known and then
run dstat before running tempest on it.
Signed-off-by: Chandan Kumar (raukadah) <chkumar@redhat.com>
Change-Id: Ia0e5218a4ff294db27b758caaa453d47945e82ec
As commented in related bug (see comments/6) we need to disable
installation of stackviz as there are issues with the constraints
used. This is not a proper fix but if it works can unblock for now
Related-Bug: 1976247
Change-Id: I8c5e168a9fd18332c073223726ad3a3e338af04d
https://review.opendev.org/c/openstack/openstack-ansible-os_tempest/+/803477
modifies the os_tempest role to create only specific tempest
resources.
Earlier private network and routers were created when network
related vars are set. Due to above patch, we are no longer creating
that.
That's why tempest_neutron_public_network_id is not getting
set and leading to failure of tempestconf generation via os_tempest.
Setting above vars fixes the issue.
Closes-Bug: #1970710
Signed-off-by: Chandan Kumar (raukadah) <chkumar@redhat.com>
Change-Id: If7318f501c72ea5fb012e2935a316eb7c3d15e5e
Set the default value to an empty string for the inventory variable
in the validation playbook.
The validation role will take of finding if there an existing inventory
on the hosts.
If no, it will fallback to localhost.
Closes-Bug: #1966184
Change-Id: I9f66c7a8992b7bb3368a4155468d3569ddf22b63
This patch adds a pre_task that includes dependency vars when
job 'dependency' is defined. This fix is needed to have 'repo-setup'
role working properly on subnodes.
Change-Id: Ie324ede9cd8cc857fbb85139ec059da427282843
The web-download test expect an url in order to work, otherwise it fails
with an error saying unable to parse the http_image.
Tempestconf set the http_image to the same as image_path if http_image
is not overwriten, so, the http_image was pointing to
/opt/cache/files/cirros. This patch set the http_image to the proper
cirros url. I check the tempest code, and only the web_download test
seems to be using the http image, so it should be safe to override the
http_image here.
Change-Id: I085177212a6e9b461009c453f7907867b118f859
Closes-Bug: #1965540
As discussed in related-bug the supplemental node used in the IPA
jobs is not using the mirror_info.sh so not using mirrors in the
repos created by repo-setup. This adds a slurp to grab the mirrors
config from the undercloud before repo-setup on supplemental.
Test and results at [1].
[1] https://review.rdoproject.org/r/c/testproject/+/40286/1#message-f825e6369163ad1f5aa98eb7c6c67b2798d5788b
Related-Bug: 1964457
Change-Id: I7b3c6f0a3b9fa7114670656df7b39cef8047fd85
This patch updates the fix to also reload NetworkManager after making
changes to its configuration file. It also adds 'ooo.test' as search
domain since it is hardcoded in deploy_freeipa.sh script.
Related-Bug: #1962727
Change-Id: If0eebb2903649bd0a9a2f4a184c2cc3daf9dc4e5
fs039 and fs064 tests are using the standard
DNS servers. This patch removes the standard
servers from the /etc/resolv.conf file and
adds the supplemental node ip.
Change-Id: I631ec520332b5db341abe9f47e7fada940ac637e
Related-Bug: #1962727
This patch add the validation.ssh_key_type into tempest.conf in order to
rum tests with FIPS enabled.
Change-Id: I045e615ce5ffce2b494263f10d27f6f046eed342
Some tests are failing because it's using the adminURL instead of
publicURL. This patch add the config option to tempestconf to force the
use of the publicURL.
Change-Id: I3d5146041df88fbcaad6b0f6d6dcb05d734251b0
Related-Bug: #1959930
CS9 jobs are failing with following error
```
'NoneType' object has no attribute 'version_tuple'
```
while installing stackviz via python_venv_build role.
The python_venv_build role uses virtualenv binary not
python -m venv utility. Moving to python -m venv might
break the role.
Till we get a proper fix, we are disabling stackviz
installation on CS9.
Related-Bug: #1953032
Signed-off-by: Chandan Kumar (raukadah) <chkumar@redhat.com>
Change-Id: Ic5bc9a5b4be36b62ce60d6db362821ea4da2239a
* Add validation playbook at the end of the ovb.yml playbook
* Make overcloud_validate playbook optional and enable by default for
backward compatibility
* Make validation playbook option and disable by default
* Use the validations.yml playbook for standalone mutlinode
Change-Id: I98726c88b5bb299b8f1fe1d49fa7a3588e051ce0