← Back to team overview

openstack team mailing list archive

A humble addition to the docs: OpenStack High Availability Guide


Hi everyone,

apologies that this took me longer than expected; I had to fight a few
fights with Maven, and resort to Anne's direct help, to get to this
point (thanks again Anne).

I've just submitted a documentation change to Gerrit that's a result
of the high-availability session we had at the Folsom Summit.
Tentatively named the OpenStack High Availability Guide, it's meant to
become a collection of reference implementations and best practices
for OpenStack HA.

It appears to us that since high availability is a topic that many
Stackers want to see as clearly optional (as in, for people who don't
want or need HA, it shouldn't be shoved down their throats), it would
be prudent to keep HA related documentation in a separate guide. This
has an added benefit, as some HA issues are clearly over-arching
several OpenStack services. For example, Nova, Quantum, Keystone,
Glance and Cinder would all rely on highly available MySQL and
RabbitMQ. Duplicating HA information for, say the Compute
Administration Guide and the Network Connectivity Administration Guide
just doesn't seem like a good idea.

So, we're starting with the following tentative outline:

* General high-availability considerations (a.k.a. "Introduction" --
elimination of SPOFs, disk redundancy, network redundancy, etc.)
* Brief introduction to the Pacemaker high-availability stack
* High availability for critical services underlying an OpenStack infrastructure
 - MySQL
 - RabbitMQ
 - possibly NFS/GlusterFS
 - Apache (for Horizon)
* High availability for OpenStack services proper
 - Keystone
 - Glance
 - Quantum
 - Nova
 - Cinder
 - Swift (only as far as explaining it's intrinsically redundant and
highly available)
* High availability for Nova instances (post-Folsom)

The change submitted today (https://review.openstack.org/#/c/8139/)
just makes the necessary addition to the build infrastructure, and
adds the introduction. We expect to be adding to this continuously
over the coming weeks and months, to the extent that our schedules
permit. :)

Some HA topics are already covered in existing documentation; we
intend to link to these from the HA guide rather than duplicate that
information. That is, unless the respective authors prefer to move
that documentation into the HA guide, but I guess everyone will want
to hold off on this for now so as not to needlessly confuse users.

Feedback, comments and contributions are most welcome.


(+ the rest of the hastexo crew)

Need help with High Availability?