Upgrade Playbooks

This section describes the playbooks that are used in the upgrade process in further detail.

Within the main scripts directory there is an upgrade-utilities directory, which contains an additional playbooks directory. These playbooks facilitate the upgrade process.


This calls a script to removes files in /etc/openstack_deploy/ansible_facts/


This playbook backs up the /etc/openstack_deploy directory before changing the configuration.

/etc/openstack_deploy copies once to /etc/openstack_deploy.LIBERTY.


This playbook ensures that the user secrets file is updated based on the example file in the main repository, making it possible to guarantee all secrets move into the upgraded environment and generate appropriately. This adds only new secrets, such as those necessary for new services or new settings added to existing services. Values set previously are not changed.


The presence of pip.conf locks down all Python installations to packages on the repo servers. If pip.conf exists on a repo server or a physical node, it will cause a circular dependency issue and the upgrade will fail.


This playbook ensures an alias is created for old hostnames that may not be RFC 1034 or 1035 compatible. Using a hostname alias allows agents to continue working in cases where the hostname is also the registered agent name. This playbook is only needed for upgrades of in-place upgrades of existing nodes or if a node is replaced or rebuilt it will be brought into the cluster using a compliant hostname.


This playbook restarts the rabbitmq nodes serially (1 at a time), and waits for rabbitmq to be back up before continuing.


The playbooks directory contains the setup-infrastructure.yml playbook. The run-upgrade.sh script calls setup-insfrastructure.yml with specific arguments to upgrade MariaDB and RabbitMQ.

For example, to run an upgrade for both components at once, run the following commands:

# openstack-ansible setup-infrastructure.yml -e 'rabbitmq_upgrade=true' \
# -e 'galera_upgrade=true'

The rabbitmq_upgrade variable tells the rabbitmq_server role to upgrade RabbitMQ.


The RabbitMQ server role installs patch releases automatically, regardless of the value of rabbitmq_upgrade. This variable only controls upgrading the major or minor versions.

Upgrading RabbitMQ in the Mitaka release is optional. The run-upgrade.sh script does not automatically upgrade it. To upgrade RabbitMQ, insert the rabbitmq_upgrade: true line into a file, such as: /etc/openstack_deploy/user_variables.yml.

The galera_upgrade variable tells the galera_server role to remove the current version of MariaDB and Galera and upgrade to the 10.x series.


Sends flush_all to memcached with the help of nc.


Update the MTU setting for networks within neutron. When migrating from Liberty to Mitaka neutron does not automatically set or migrate networks. Neutron has no migration to correctly set the MTU on existing networks created in Liberty or below. This work-around sets the MTU on these networks before the upgrade starts so that instances booted on these networks after the upgrade has finished will have their MTUs set correctly.


This playbook ensures the OpenStack service databases for heat, nova, neutron, and cinder are cleaned up with regard to actively registered services. These OpenStack services register nodes within the environment using the hostname as a unique key and in previous releases containers had invalid hostnames. This playbook will remove invalid service/agent entries that are found within the database and meet the invalid hostname criteria.