[ English | Deutsch | русский | español | Indonesia | English (United Kingdom) ]


What is the OSA release model

OpenStack-Ansible (OSA) uses the ‘cycle-trailing’ release model as specified in the OpenStack release model reference_.

How are release tags decided?

In order to ensure a common understanding of what release versions mean, we use Semantic Versioning 2.0.0_ for versioning as a basis. The exception to the rule is for milestone releases during a development cycle, where releases are tagged <MAJOR>.0.0.0b<MILESTONE> where <MAJOR> is the next major release number, and <MILESTONE> is the milestone number.

The OpenStack series names are alphabetical, with each letter matched to a number (eg: Austin = 1, Bexar = 2, Newton = 14, Pike = 16, etc). OSA adopted the same <MAJOR> release numbering as the Nova project to match the overall OpenStack series version numbering.

How frequently does OSA release?

Major releases are done every six months according to the OpenStack release schedule_. Each major release is consistent with an OpenStack series.

Minor/patch releases are requested for stable branches on the second and last Friday of every month. The releases are typically completed within a few days of the request.

What version of OpenStack is deployed by OSA?

For each OSA release, the OpenStack version that is deployed is set to a specific OpenStack git SHA-1 hash_ (SHA). These are updated after every OSA release. The intent is to ensure that OSA users are able to enjoy an updated OpenStack environment with smaller increments of change than the typical upstream service releases allow for as they are usually very infrequent.

This does mean that a stable OSA deployment will include a version of a service (eg: nova-17.0.3dev4) which does not match a tag exactly as you may expect (eg: nova-17.0.3).

If you wish to change the SHA to a specific SHA/tag/branch, or wish to use your own fork of an OpenStack service, please see the section titled Overriding other upstream projects source code in the user guide.

When does a patch to an OSA role get into a release?

For each OSA release, the Ansible roles that form that release are set to a specific git SHA-1 hash_ (SHA). These are updated after every OSA release.

OSA frequently does pro-active bugfix backports. In order to reduce the risk of these backports introducing any destabilization, OSA implements a ‘soak’ period for any patches implemented in the stable branches for roles, but also provides for circumventing this in exceptional circumstances.

A patch merged into a role is immediately tested by other role tests, ensuring that any major breaking change is caught. Once a minor/patch release is requested, the integrated build receives a ‘SHA bump’ patch to update the integrated build to using the latest available roles including that new patch. This new set is available for testing to anyone wanting to use the head of the stable branch, and is tested in periodic tests until the next release. In total, that means that the cycle time for a patch from merge to release is anywhere from two weeks to one month.

If there is a requirement to rush a role patch into the next release, then anyone may propose a change to the ansible-role-requirements.yml file in the openstack/openstack-ansible repository with the appropriate justification.

We believe that this approach brings a balance of both reasonable stability, while still being able to do pro-active backports.

The only exception to this rule is for the master branch, which intentionally consumes the master branch from all roles between releases so that any changes are immediately integration tested.