Failed containers are pretty bad, since we have a degraded service.
Running this validation before an upgrade is a good thing, and
running it after a deploy/upgrade will ensure we're in a right state
Co-Authored-by: Gaël Chamoulaud <gchamoul@redhat.com>
Change-Id: I242f1c7cff76e8304696ea10b32c1545fa5b8ea5
This [1] introduced support tripleo-ansible-inventory for multiple
stacks/plans and now configs.plan needs to be list/iterable.
[1]: I23fdeaa4950244bf3aec8e8ea71e8a29f763c074
Change-Id: I929a3490d88602eaf62bbe1a559febfc185aa7e9
The 'neutron-sanity-check role could be executed against the UC but also
against the Controller nodes. However, the 'container_cli' variable comes
from two different sources of information. This roles should manage this.
OVN is now the default ML2 plugin for the OC nodes while the UC is still
using OVS. Then, this role has to run the 'neutron-sanity-check' command
on two different containers as well:
- Undercloud -> neutron_ovs_agent
- Controller -> neutron_api
This patches fixes both of these issues.
Change-Id: I3a027282243db1645361e987cc63f7b6f6fc78ce
Signed-off-by: Gael Chamoulaud <gchamoul@redhat.com>
Due to the Panko deprecation for the Train release.
Change-Id: Ic6b497b033f52342eb9dcdbdec84c406efecb5ac
Signed-off-by: Gael Chamoulaud <gchamoul@redhat.com>
If tripleo-ansible-inventory is called with '--stack foo,bar',
then build an inventory for stack foo and stack bar; then
merge the two inventories into one before writing it out to
an inventory file.
An ansible inventory host group will be created for every
item in the cross product of stacks and roles. For example,
central_Controller, edge0_Compute, edge1_Compute, etc.
Change-Id: I23fdeaa4950244bf3aec8e8ea71e8a29f763c074
Depends-On: I522f1acbd39ad36ae1fec896188d3e20e0575566
Implements: blueprint split-control-plane-unified-inventory
The parameter undercloud_connection is being available as
an input option in the script, however it was not being passed
into the TripleoInventory object instantiation.
Change-Id: Ife819b1d807946c8e53deaacbb8e4f9361009f6e
Closes-Bug: #1832932
This patch fixes the right default for the 'undercloud-connection' option
according to the help message.
Signed-off-by: Gael Chamoulaud <gchamoul@redhat.com>
Co-Authored-By: Jose Luis Franco <jfrancoa@redhat.com>
Change-Id: Ib0eb19ddd2968ab328565dec3171f480327be27f
To enable the execution via ssh into the Undercloud node it's
required to pass also the ansible_ssh_user private key. This
options is part of the tripleo_common inventory.py, but defaults
to None.
This patch allows to pass this parameter in case it's needed.
Also, this patch allows to pass a different value for the host_network
parameter in TripleoInventory instantiation. Setting it by default to
ctlplane.
Change-Id: Id303f356c1b06906aa5aae43b0032bb80a712508
Related-Bug: #1832932
We can't easily add host variables, meaning we have to modify the
validation in order to accept external variable.
The Molecule test validate it's still working as expected.
Change-Id: I81edce1f949c0ad38d4ccef2a94a45f455cd9a9c
We now use a promote job for publishing that avoids building again the
documents after merge. use this here as well.
We can only use the promote jobs when the job run in gate, so use the
same file conditions. To make those easier to sync, get rid of the
special tripleo job and use openstack-tox-docs directly.
For further details about promote jobs, see
http://lists.openstack.org/pipermail/openstack-discuss/2019-August/008610.html
Change-Id: I8432eab9fcec488dc757eabc350ea3e6a9c53102
As of recently, deploy images for overcloud nodes may now be file-based
rather than in Glance. And after upcoming changes to Ironic,
driver_info.deploy_kernel and driver_info.deploy_ramdisk may be empty.
With the need to be both forwards- and backwards-compatible, the
validation logic here has been necessarily loosened: merely checking
that nodes of the same arch/platform do not have overly diverse deploy
images associated. Such validation is not as entirely definitive or
detailed as its predecessor, but also far less convoluted than any
alternative...
Change-Id: I5aa4e98770f29cd2df3f39e4000b822681c551f4
This patch does two things:
- ensure we don't rely on systemd within the validation itself
- adds molecule tests
Since molecule tests are running in docker containers, we can't easily
use the "systemd" helpers.
This patch makes sure the validation checks if the port is opened, and
if we get a proper answer from the server.
Change-Id: Id9a1a62e0cfff0b5300af41ec6c7a08f05ddf012
This patch also transforms the Nova configuration file path into an
Ansible variable.
Change-Id: I0bbbe16fbb21acef79a5f66fbea963ce3fd0f57d
Signed-off-by: Gael Chamoulaud <gchamoul@redhat.com>
This change implements a native zuul runner for all role tests.
The change will now run role tests, executing molecule, in the
same was as tripleo-ansible.
To ensure we're passing lint checks, several files have been
updated to resolve lint issues which are now more strict due
to the pre-commit changes.
Scripts have been added to allow developers to run molecule tests
on local environments. These tools will allow developers to mimic
the upstream test process locally.
Change-Id: I07bbcc0b331aa89dafdae5978ea2bb4859a59143
Signed-off-by: Gael Chamoulaud <gchamoul@redhat.com>
- Adds bashate test using v0.6.0
- Bump pre-commit-hooks release to v2.2.4
- Bump ansible-lint release to v4.1.0a0
- Fix some minor flake8 errors
Change-Id: I66b796fab1d8651163226febbc4e99648a9ecc6a
Signed-off-by: Gael Chamoulaud <gchamoul@redhat.com>
This patch adds automation to add documentation for each Ansible roles
and custom modules for TripleO-Validations.
Patch-Inspired-By: tripleo-ansible-Team++
Change-Id: I2c8ec698cdcbc304b5d9f3bd72bcede5bf2aafd3
Signed-off-by: Gael Chamoulaud <gchamoul@redhat.com>
When the Ceph cluster name is customized the ceph commands need
to be told the custom cluster name.
Change-Id: Ic17307efc2b4907cc21df8289a2e0c84d2f30b48
Cédric Jeanneret