Chapter 1. Overview

The sample HA deployment used for this document was created using the following guides as reference:

] shows the particular configuration that was built specifically to test the high availability features described here. For details on how to recreate this setup so you can try the steps yourself, refer to xref:buildenv[.

Figure 1.1. OpenStack HA environment deployed through director

In an HA deployment, all OpenStack services must be launched and managed by either Pacemaker or HAProxy. This includes all related and dependent services.

For example, the httpd service is required by openstack-dashboard. As such, in an HA environment, httpd must not be launched or enabled manually (eg. through systemctl rather than pcs). Many colocation or dependency problems in HA deployments are caused by services being managed outside of Pacemaker or HAProxy.

To avoid this, orchestrate your HA deployment entirely in director. The templates and puppet modules used by director ensure that all services are configured and launched correctly, particularly for HA. Further, when troubleshooting HA issues, always interact with services through the HA framework whenever possible.