astack.sh

How astack.sh deploys a Cloud

The astack.sh script drives a number of other scripts to setup the specified cloud, i.e. the following phases:

Setup

This phase can be skipped on subsequent runs using the --no-setup option to astack.sh

• Sets up system environment (by running ardana-dev-tools/ansible/dev-env-install.yml playbook).
  • Installs Vagrant
  • Installs & sets up kvm & libvirt
  • Creates an ansible 1.9 based venv to use when running ansible-playbook operations. Make sure to run source ardana-dev-tools/tools/venvs/ansible/bin/activate if you are running playbooks directly.

Build

This phase can be skipped on subsequent runs using the --no-build option to astack.sh

• Implemented in the ardana-dev-tools/bin/build-venv.sh script
• Will trigger some of setup operations if needed, on first run
  • Ensures most required artifacts have been downloaded
    • Some may be downloaded later if needed
  • Creates Vagrant boxes for specified platforms
• Clones (or refreshes) the git repos for the sources backing the venvs
  • Implemented in the ardana-dev-tools/ansible/get-venv-sources.yml playbook
  • Uses a git server corresponding to ARDANA_SITE (see above)
  • NOTE: Can be skipped by setting ARDANA_GIT_UPDATE=no in environment, or by specifying --no-checkout when running build-venv.sh directly.
• Brings up, using Vagrant, required build VMs to build the venvs needed to deploy a cloud
• Runs the ardana-dev-tools/ansible/venv-build.yml playbook to build venvs.

Cloud Provision

• Bring up the specified cloud, using Vagrant
  • Uses the servers.yml file in the associated cloud definition (e.g. ./scratch-master/input-model/ardana-input-model/2.0/hp-ci/deployerincloud-lite/data/servers.yml) to specify the VMs in the cloud
• During provisioning of the deployer VM (which may not be called deployer) will trigger the creation of the product tarball (FIXME: this isn't part of the Build phase for historical reasons; that technical debt will probably have to be addressed as we move the Build phase into OBS)
  • Contains ansible sources
    • Clones (or refreshes) the ansible sources git repos using ardana-dev-tools/ansible/get-ansible-sources.yml playbook
    • Can be skipped by setting ARDANA_GET_UPDATE=no in environment
  • Contains venv repos for each platform, including manifests descripting venvs
  • Contains generated ardana-init.bash script used to install product tarball bits and setup deployer environment
  • NOTE: Can be skipped by specifying --tarball ${absolute_path_to_deployer_tarball} when running astack.sh
• Uploads the product tarball to the deployer node and unpacks it to /home/$ARDANAUSER
• Runs the ardana-init.bash script from the unpacked product tarball
  • Sets up venv repos
  • Sets up ~/openstack git repo and other product areas under the deployment account on the deployer
  • Deletes itself and the other artifacts from the unpacking

Cloud Configure

This phase can be skipped on subsequent runs using the --no-config option to astack.sh.

• Copies specified cloud model from ardana-input-model/2.0/hp-ci to ~/openstack/my_cloud/definition on deployer and commits it into the ~/openstack git repo.
• Configures the specified cloud by running the config-processor-run.yml playbook from the ~/openstack/ardana/ansible directory on the deployer to compile the specified cloud definition
• Sets up deployment scratch area by running the ready-deployment.yml playbook from the ~/openstack/ardana/ansible directory on the deployer to construct ~/scratch/ansible/next/ardana/ansible.
  • The scratch area consists of the ansible sources found under ~/openstack/ardana/ansible and the configuration processor generated hosts/verb_hosts file and host_vars and p_vars` hierarchies.

Cloud Deploy

This phase can be skipped on subsequent runs using the --no-site option to astack.sh.

• Runs the site.yml playbook from the ~/scratch/ansible/next/ardana/ansible area to deploy the specified cloud.