By default tripleo doesn't enable cinder-backup service, this is done in
scenario002, so these tests must be in the skip file otherwise it will
always fail.
Change-Id: I54c80606e7f761843782a5eafd0bfa6045ab2c73
Sometimes we want to collect everything, but specific files and
folders. Support exclude list for this.
Change-Id: I10302ee50c5539fbacd539371dde0e7d0d7c4f71
This commit manages all the stdout and stderr status of each HA
validation test.
In this way it will be possible to have all the logs related to the
tests in one place (on the undercloud) that will be collected by the
collect_logs role and so accessible from the external location where
the logs will be stored.
Change-Id: Ie803a920a17bd2f24ccd89f733781dd7e4d07fdd
With the Newton and later releases, it is possible to introspect
only nodes in the manageable state rather than run bulk
introspection that would work on all imported nodes.
Change-Id: Ic36dea3f93e84ab49731996a69a9d3c28f973e86
Pip isn't installed on undercloud, making Install stackviz task fails
because isn't able to find it.
Change-Id: I3aac9a91e1ca74f03d84694239b0a8299dfc9c85
We need a simple, togglable way to verify that Sphinx has, at a
minimum, generated each of the files outlined within the
`artcl_create_docs_payload.table_of_contents` and added them to
the resulting index.html
- Add togglable var to collect-logs role, `artcl_verify_sphinx_build`
default to false
- Update artcl_collect_dir to pull from common/defaults
Change-Id: I76fef28a026730818d44274ab01741034f9fa40d
This will now run bashate on all shell scripts. We could ignore the
following list of bashate errors:
- E006: Line longer than 79 columns (as many scripts use jinja
templating, this is very difficult)
- E040: Syntax error determined using `bash -n` (as many scripts
use jinja templating, this will often fail and the syntax
error will be discovered in execution anyway)
Change-Id: I6f6d454ac54a0ac98d8b8a092fa522cba092a592
Signed-off-by: Gael Chamoulaud <gchamoul@redhat.com>
Previously the 'flake8' command was running but didn't seem to
check out something in all cases.
- This patch decouples the 'ansible-lint' and the 'flake8' sections
into separated 'testenv'.
- And this patch fixes all the 'flake8' issues.
Change-Id: I30169598d086515121175ca322882ee00f05b913
Signed-off-by: Gael Chamoulaud <gchamoul@redhat.com>
novaclient 7.0.0 and above requires the 'project_name'
parameter, defined in the cloud.yml config file.
Change-Id: I15928d05ddbad28a8bdb3ef247c6aac85226a4d2
When publishing logs in usptream we need an option to get
all files from host together both getting only one of them.
Save and upload the whole tar.gz for host together with flattened
logs and configs.
Change-Id: Ib272841a9ddf15d9ccd36a5d42a9dd94bf934f8f
* silence the console output from curl but keep error output to avoid
garbled download output in the logs
* acquire the floating IP address from the stack output to test the same
way as tripleo-ci
* follow up the heat template differences (key -> key_pair)
Change-Id: I4c2bbbc3d542bf62f375d81f70069dff2490fc36
Remove unused undercloud_post_install_script and log from parameters
and documentation since it's not used anymore.
Change-Id: I88de9555b17d83b1359e2d7ca88a99818f3e07a1
This commit defines a var file for each release containing the list of
the HA validation tests that should be done. These new variables are now
also used as booleans to determine if a test should be executed or not.
With this new approach a user will be able to exclude (or include) tests
at runtime, like this:
./quickstart.sh \
--retain-inventory \
...
...
--release mitaka \
--tags "overcloud-validate-ha" \
--extra-vars test_ha_failed_actions=false \
--extra-vars test_ha_master_slave=false \
--extra-vars test_ha_keystone_constraint_removal=false \
--extra-vars test_ha_instance=true \
$VIRTHOST
In this case I will test just the instance deployment.
This means that without any extra var tests will be the default
depending on the release, otherwise specific tests will be made.
It also updates the README with all these new informations.
Change-Id: I113b8e929dddb9ff18b57faa905a754e2f6b5999
If the task or role is skipped, ansible still tries to evaluate
variables and doesn't consider set_fact, having ansible change
list as undefined, which stops playbook with fatal error.
Change-Id: I7417918b26893fb37bed35260eee7cc98feaea9a
This commit fixes the way the test suite is cloned on each node.
With the new ansible release the old fashion is not good anymore, it
fails while substituting items.
This solution has been tested and it is working.
Change-Id: I7609945b5ef2e6b932ba23441c7e773631dc60b4
The current implementation of build-test-packages allows for
precisely one method of providing what changes to build.
Either passing a artg_change_list dict directly via variable,
reading GERRIT vars from the env, or reading ZUUL vars from
the env.
This patch makes it possible for multiple of these methods to
be used without stomping on eachother. It is needed to be able
to run the gate which specifically exercises this role, but may
be useful in other scenarios as well.
Change-Id: Ib2290411865f9bb1a2804af64e85541530ae2e97
Jenkins