add the ansible variable options to turn on a
add the pre-deploy tasks to setup the yaml files
required by the containerized compute
In an effort to make more of the tripleo deployment ci more composable
it has been discussed to break out the undercloud deployment into it's
own role. There are examples where additional configuration is needed
prior to the undercloud installation such as dpdk, and installing in
other ci environments.
The overcloud-deploy.sh script first spins while waiting for hypervisor
resources to become available. In deployed-server environments this will
never be true, causing the job to spin indefinitely. This commit makes
this an optional component with a hypervisor_wait variable.
This commit refers to the baremetal undercloud role:
- Fix baremetal undercloud playbooks to support new extra roles names;
- Define the repos as a list which can be overridden;
- Removes the restart of libvirt before the customization of the
overcloud image. This was initially included because it seemed to be
necessary for virt-customize to work, but further tests confirmed it is
not needed anymore;
This error seems to be environment/workflow specific.
The workaround was added as the error is not
reproducible in other environments - and those
not using ironic node-update.
Uploading the logs can take too much time, resulting ssh connection
timeout during the ansible task execution. Log collection usually runs
in post build tasks, making it immune to jenkins job timeouts.
This change makes the upload tasks asyncronous and adds a timeout.
- Could be situation when nothing is built and it's OK, i.e.
- Use zuul host of openstack by default and dependency on zuul
changes, not host (because no ZUUL_HOST is set in upstream)
- Build only once for each project (as in current CI do)
We should run and publish a minimal sets of files even without any host,
when we failed the run before inventory generation.
This change separates the collection step that runs on all hosts except
localhost, and the rest running on localhost. Running on localhost
always succeeds, even with an empty inventory.
Also add a log environment file for local collect-logs.sh runs that
does not upload logs.
* remove the roles build function and relevant documentation that moved
to the gate-quickstart role
* follow up the role rename in README.md
* add tripleo-ci to the skipped projects
* remove empty/placeholder files
* add and rename the playbook build-test-packages.yml used by the
Zuul uses different variables and files to handle artifact uploading,
this change adds the required modifications to collect and publish logs
in the rdoproject.org Zuul environment.
https://review.openstack.org/#/c/397312/ has removed some options in
Specifically, --neutron-network-type and --neutron-tunnel-types
are being used in tripleo-quickstart .This patch removes them
from extra_args and scale_extra_args.
This parameters are being passed via environment files or using vxlan
by default so i think it doesn't need any other adjustment.
- Remove scale_ceph.yml config, it isn't tested/verified and
shouldn't be supported
- Clean up the scale_compute.yml config and remove unused or
redundant (overriding default with default) variables
- Ensure the conditional tasks cast strings to bools to avoid
If we want to use the build-test-packages role to gate on DLRN
changes itself, we need it not to clone DLRN if it was already done
by Zuul. Adding a check to skip it if needed.
For environments where OVB is run on an Openstack
host cloud, this role contains the functionality
to manage the heat stack.
The associated playbooks are included.