Summary

Dealing with scale is a very important task in enterprise deployments. As your number of nodes increases beyond the proof-of-concept stage (> 50 nodes), the simple WEBrick server cannot be used. In the first section, we configured a Puppet master with passenger to handle a larger load. We then expanded that configuration with load balancing and proxying techniques realizing that Puppet is simply a web service. Understanding how nodes request files, catalogs, and certificates allows you to modify the configuration and bypass or alleviate bottlenecks.

In the last section, we explored masterless configuration, wherein instead of checking into Puppet to retrieve new code, the nodes check out the code first and then run against it on a schedule.

Now that we have dealt with the load issue, we need to turn our attention to managing the modules to be applied to nodes. We will cover organizing the nodes in the next chapter.