It's easy to set up a single GNU/Linux server, but there's no obvious way to take those same skills and build functional clusters. If you want to build and maintain clusters while avoiding the pitfalls of one-by-one server management, you need a whole new class of tools. Unfortunately, documentation for choosing, deploying, and integrating these tools is sparse.

At Four Kitchens, we've done the hard work: we maintain an in-office lab to develop and apply cluster-level integration and management techniques to the challenging problems we encounter managing our clients' infrastructure.

This session will focus on taking a set of servers with little more than operating systems and SSH and building a cluster you can scale, monitor, and maintain.

Session attendees should be familiar with single-server Linux package management and configuration.

Learn how to:

• Run the latest stable versions of MySQL 5.0 and PHP 5.2 on CentOS 5
• Install APC and memcached
• Deploy your code (core, modules, etc.) to sets of servers from a version control system (Bazaar/Subversion and Capistrano)
• Maintain consistent configurations and automate deployments (bcfg2)
• Monitor servers for load and functionality (Nagios)
• Graph historical system metrics (Munin)
• Automate fail-over (Heartbeat 2)

And we'll do it all with easy-to-update yum repositories and no manual compilation or installation.

I would like feedback from prospective attendees on the most important items above. It's unlikely I'll be able to give comprehensive coverage to all of them.

Note: The actual tools used to implement the items above may change. We probably won't use all the same tools by the time Drupalcon comes around.