Yatin Karel
581d0a1d60
Currently on the multinode jobs, job continue to run even if devstack setup fails on any of the subnode and then fails later when required conditions are not met. This patch changes it to fail if any of the node setup fails using any_errors_fatal: true. Change-Id: I2acd8a1fe0802ee1880df2ef794f8e7d7478b67b Signed-off-by: Yatin Karel <ykarel@redhat.com>
Orchestrate a devstack
Runs devstack in a multinode scenario, with one controller node and a group of subnodes.
The reason for this role is so that jobs in other repository may run devstack in their plays with no need for re-implementing the orchestration logic.
The "run-devstack" role is available to run devstack with no orchestration.
This role sets up the controller and CA first, it then pushes CA data to sub-nodes and run devstack there. The only requirement for this role is for the controller inventory_hostname to be "controller" and for all sub-nodes to be defined in a group called "subnode".
This role needs to be invoked from a playbook that uses a "linear" strategy.
Role Variables