Improve the High Availability Guide¶

Problem description¶

Presently, the High Availability (HA) Guide is incomplete. Information is sparse and there are sections that have simply not been filled in.

The current guide is also out of date with regards to current best practices, and it contains unnecessary information that should be removed.

Proposed change¶

Firstly, reach consensus on the intended audience and high-level use cases for the guide:

• As a cloud deployer, I need an OpenStack HA guide so that I can understand both the architectural principles behind building an HA OpenStack cloud, and the concrete steps involved in an implementation.
• As a cloud operator, I need an OpenStack HA guide so that I can understand how an existing HA OpenStack cloud works and what is required for its maintenance.

Based on that consensus, this spec proposes that the HA guide aims to define, justify, and explain a high level architecture for a highly available setup which uses the Pacemaker cluster manager to provide:

• Detection and recovery of machine and application-level failures
• Startup/shutdown ordering between applications
• Preferences for other applications that must/must-not run on the same machine
• Provably correct response to any failure or cluster state

The guide will aim to stay relevant across all distributions whilst not attempting to give every tiny detail about how to implement HA for each distribution. It will also aim to avoid duplicating too much information which can be found elsewhere. For example, basic package installation for a given distribution.

Since the existing guide already has plenty of helpful and relevant information, the proposal for this guide aims to avoid any rip-and-replace action preferring instead incremental change.

Andrew Beekhof (specialty team lead) proposes to use the following document as a reference providing updated information for the improved guide: https://github.com/beekhof/osp-ha-deploy/blob/master/ha-openstack.md

Note

The above Github document contains heavy Red Hat content. Some may be included in the final publication of the HA guide however it will be structured such that advocates of other distributions/tools will be able to easily add alternatives.

Implementation¶

Dependencies¶

• Potentially dependent on community engagement and SMEs providing content.

Testing¶

This document will be tested by the community as it is updated.

References¶

1. Mailing list discussion, December 2016: http://lists.openstack.org/pipermail/openstack-docs/2016-December/009410.html
2. Growing list of relevant bugs: https://bugs.launchpad.net/openstack-manuals/+bugs?field.tag=ha-guide