server_patterns

Pattern: Start with a simple standalone server and worry about scaling later.

Discussion: Many organization bottleneck their Chef adoption by solving a problem they don't have: scale and availibility. First, 'high-availability' is not largely necessary ... etc.

Antipattern: Building a chef-server cluster that includes all the nodes in all the datacenters

Discussion: The issues with latency, reliability and scale will overwhelm the gain from having search and policy absolutely consistent globally. Search should be replaced with DNS or equivalent.